[agi] Russel: If you can figure out another way to do it, I'm all ears!
Russel Said: *Oh, I can figure out how to solve most specific problems. From an AGI point of view, however, that leaves the question of how those individual solutions are going to serve as sources of knowledge for a system, rather than separate specific programs. My answer is to build something that can reason about code, for which formal logic is a necessary ingredient. If you can figure out another way to do it, I'm all ears! *Well, there are at least two problems here. *1) How to gain initial knowledge 2) How to use knowledge to achieve goals once we have it. * *1) How to gain initial knowledge* Ah, this is something very cool that I've been working on lately. Pick a particular example of initial knowledge from the example below and we can trace how it is learned and how such learning mechanisms can be implemented. There are many, so I'm not going to try to list them. I thought it would also be more fun for you all to pick one and surprise me. *Let's start with a simple example of 2 (using knowledge we already have and learning more) : Creating a Hello World program* Note that many of the details in how the reasoning is done are left out because 1) they are yet to be determined in detail and 2) the email is long enough without them. *Initial Assumptions: * The agent has some initial knowledge about programs, where one might find information about programming. The agent might have a text book on it. The agent understands what a hello world program is supposed to do. So, what are we solving for if the agent has so many initial capabilities? We're trying to show how the agent reasons about what it already knows to achieve a goal. The goal is to create a program that says hello world. The agent understands this by reasons about statements made in a textbook about the hello world example program. The agent has to plan its actions to achieve the intention write a hello world program. The plan is not a complete step by step plan. It just tells the general direction to go. This is the rough to fine heuristic that human beings often use. From there, does mean's ends analysis, searches for and finds information that might be relevant to the situation at hand, and reasons about what they've done in the past that have help achieve parts of such a goal. The AGI knows that programs can be created through the visual studio's IDE, based on reading about programming in C# (the book he/she has). So, it realizes that it needs to achieve a subgoal of finding visual studio's IDE to use it. It knows it can do this by getting to the computer and clicking on the icon that it knows is associated with visual studio. The program comes up. So, then we ask ourselves what's the next step?. Our brain has marked memories associated with creating programs. It has recorded the fact that we clicked on the file menu to create a new program and that this was part of the process in achieving the goal. So, our memory pulls this fact and executes the action because we have no reasons to not pursue the action in memory. So, to this we go to the file menu and click create a new project. We also pull in relevant information, which says you have to do this that and the other also if we want to create a program. We pull in relevant info from what we read in the text book about what to be careful of and what has to be done, etc. What's next? We want to make the program print out hello world. we recall that we can do this by using the command Console.WriteLine(). and we recall that the thing printed out was in between the parantheses like so: Console.WriteLine(something to print out); So, we hypothesize that if replace what was printed out with hello world that it will work. so we try Console.WriteLine(hello world). it works! hurray. toda. done. Yeah, I know. It's over simplified. But you can see the types of reasoning that are required to achieve such a task. Do this thought experiment on enough problems and generalize what it takes to achieve them (don't try to overgeneralize though!). DO NOT THROW OUT the requirements. You cannot throw out computer vision because you don't know how to implement it. Sensory perception is a requirement for AGI for many reasons. So, just make it an assumption in your design until you can work out the details. We'll do the same thought experiment on computer vision as well to see how it can be integrated with the whole system. For now though, we're just focusing on this simple programming task. * * --- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/8660244-d750797a Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] Very Cool Object Name Intent Test
I just came up with an awesome test. Ask someone, anyone you know to name something really big and obvious around them that they already know the position of. Tell them to point to it and name it. Practically *every* time, they will look at it just before or as they are naming it! And it feels incredibly uncomfortable not to look at what you are naming as you are trying to communicate that. These are the sorts of built in cues that children require to learn language. The children know when they are being addressed, and they know how to narrow the possible things that you intend to refer to when talking to them. Pointing gestures, eye movements, etc. They all are very strong *tells* (like in poker) regarding the intent of your speech. We are constantly analyzing the actual intent of speakers and then interpreting what they say. This is how children and adults learn language and gain experience :) I'm working on a rough to fine model of this in my Pseudo AGI design. --- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] Pseudo Design as a Solution to AGI Design
I've been to think lately that the solution to creating a realistic AGI design is psuedo design. What do I mean? Not simulation... not practical applications... not extremely detailed implementations. The design would start at a high level and go deeper into detail as possible. So, why would this be a solution? Well, before I mention the cons to this approach, consider the following: *Problems it would solve:* 1) There is no money and little interest for AGI. Even if you could get money, I am 99.99% sure it would be spent wrong. I know, I know... I'm supposed to be trying to get us money, not dissuade it. But, I really think we are repeating the mistakes of earlier researchers that promised too much on unjustified ideas. Then when they failed, it created AI winters, over and over and over again. History repeats itself. So, getting us more money would likely do harm in addition to too little good, the way it would be spent, for me to care. Extremely few people are interested in AGI and among those that are, their ideas about it are very, very flawed. We tend to approach the problem using our typical heuristics and problem solving techniques, but the problem is no longer amenable to these techniques. For example, the idea that patterns finding is sufficient for intelligence. It has not been proven beyond my reasonable arguments against it. Yet, people are getting funding and pursuing entire architectures based on it. Does that really make sense? Nope. We must pseudo test and pseudo design our algorithms first. Why? Because after spending several years on these designs that I can reasonably predict will fail with a high likelihood, we'll be back as the same place we were. Wouldn't we be much better off figuring that out earlier rather than later through fast prototyping techniques, such as the one I mentioned (pseudo design and testing)? 2) Implementations tend to get overwhelmed by the desire to show immediate results or achieve practical short-term goals. This completely throws off AGI implementations, because these other constraints are not compatible with more important AGI constraints. 3) We could find a solution much faster... AGI is a massively constrained CSP (Constraint Satisfaction Problem). The eternity puzzle is a great example of such a problem. If you approach the eternity puzzle using heuristics alone to generate a likely solution, such as how pretty the pattern is, or how plausible it is that the designers created this design, it is guaranteed to fail. This is especially true if it takes you even a few minutes to reject the design. The puzzle has so many possibilities that if you were to try to look at each one to see if it was a solution, it would literally take an eternity. So, how do you solve such problems? You start with the most constrained parts of the puzzle first, and you use heuristics to guide your search for solutions paths that are likely to contain a solution and avoid solutions paths that are less likely to contain a solution. Most importantly, you have to try a lot of solutions and reject the bad ones quickly, so that you can get to the right one. How does this apply to AGI? It's almost exactly the same. Current researchers are spending a lot of time on solutions that were generated using bad heuristics (unjustifiable human reasoning heuristics). Then they take forever to test them out (years) before they inevitably fail. A better way is to test solutions with as minimal effort and time as possible, such as by using pseudo design and testing techniques. This way you can settle onto the right solution path much, much faster and not waste time on a solution that clearly wouldn't work if you simply spent a bit more time analyzing it. Yes, such an approach has problems also, such as dishonesty or delusion in how the algorithms would actually work. I'll mention these more below. But, we have those delusions and problems already :) So, overall, this approach seems to be significantly better. 4) if we could show that a pseudo AGI design works in sufficient detail and with sufficient plausibility, it would likely change the minds of: -many people that don't think AGI is possible, -those that think it isn't possible in their lifetimes, and -those that think it isn't worth investing in. In other words... we would get the money, help and interest needed to make it happen. Demos are great at generating interest in things that are very complicated. This would be a fantastic demonstration. *Pros:* 1) Fast design testing and rejection 2) Rough to fine design... would arrive at a solution faster because it uses the *Most*-*Constrained*-Variable-First heuristic (such as has been used to solve the eternity puzzle... you solve the most constrained portion first to avoid having to try out many possibilities that will fail at the most constrained part). 3) Less pressure for practical applications and more focus on important AGI issues... this removes extra constraints that are not
[agi] Wow.... just wow. (Adaptive AI)
I accidentally stumbled upon the website of Adaptive AI. I must say, it is by FAR the best AGI approach and design I have ever seen. As I'm read it today and yesterday (haven't quite finished it all), I agreed with so much of what he wrote that I could almost swear that I wrote it myself. He even uses the key word I've begun to use myself, which is explicit AGI design. This dude is awesome. If you haven't read about it yet, please do: http://www.adaptiveai.com/research/index.htm Dave PS: I don't agree with absolutely everything per say, such as the fuzzy pattern matching stuff... because I just don't understand the specifics, pros and cons of it to agree or disagree. But, damn, this guy got enough of it right that I have to applaud him regardless of the other details. --- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] Human Reasoning Examples
Does anyone know of a list, book or links about human reasoning examples? I'm having such a hard time finding info on this. I don't want to have to create all the examples myself. but I don't know where to look. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] Alternative way to reverse engineer the brain
Has anyone thought about sort of self-assembling nano electrodes or other nano detectors that could probe the vast majority of neurons and important structures in a very small brain (such as a gnat brain or a C. Elegans worm, or even a larger animal)? It seems to me that this would be a hell of a lot easier than simulating a brain, since there are waay too many factors and dynamics involved to get the simulation to be accurate. Maybe we could just invent a way to probe every part of the brain in vivo. Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] Language Acquisition TV Special
I've become extremely fascinated with language acquisition. I am convinced that we can tease out the algorithms that children use to learn language from observations like the ones seen in the video link below. I'm about to start watching the second video, but thought you guys might like watching this too :) Check it out! Also, if you haven't done so yet, check out William O'Grady's book How Children Learn Language. I love that book. http://www.youtube.com/watch?v=PZatrvNDOiENR=1 Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] Neuroplasticity Explanation Hypothesis
I just had this really interesting idea about neuroplasticity as I'm sitting here listening to a speeches at the Singularity Summit. I was trying to figure out how neuroplasticity works and why the hell is it that the brain can find the same patterns in input from completely different senses. For example, if born without eyes, we can see with touch. If born without hearing and vision, we can also see and hear with touch! (an example of this is a blind and deaf person putting their hand on your mouth and neck to detect and understand your speech. this is a real example). How the hell does the brain do that?! The brain knows how to process certain inputs just the right way. For example, it knows to group things by color or that faces have certain special meanings. How does it know to process this sensory input the right way? I don't think it's purely pattern recognition. Actually, it cannot be just pattern recognition alone. So, I realized that it would make sense that cells don't create a network and wait for input. The cells are not specialized *before* they get sensory inputs or other types of input (such as input from nearby cells). These cells specialize AFTER receiving input! That means that our DNA defines what patterns we should look for and how to process those patterns. Guess what that means! That means that if these patterns come from completely different sensory organs, the brain can still recognize the patterns and the cells that receive these patterns can specialize just right to process them a certain way! That would perfectly (so I believe) explain neuroplasticity. Basically, it is a side-effect of the specific design of our brains. But, it means that the brain is not just a pattern recognizer. It has built-in knowledge which is absolutely essential to process inputs correctly. This supports my hypothesis that artificial neural nets are not correctly design to be able to achieve AGI the way the brain does. This would also explain my beliefs that the brain knows how to process in ways that correctly represent true real-world relationships. It would also explain why this processing can self assemble correctly. The knowledge for how to process inputs is built in(my hypothesis), but it self assembles only when inputs that have certain patterns and chemical signals are presented to the cells. This would explain the confusion for between purely self-assembling models and built-in knowledge of how certain patterns or input should be processed. Clearly, the brain does not evolve to process world input correctly every single time a person is born. We solved this problem already through our DNA and billions of years of evolution. So, the solutions to the problems are built into our DNA. This would also explain how the brain is able to handle other important functions such as: memory, hierarchical relationships, etc. When the brain detects the need and the right patterns of specialized cells, it can then create even more specialized cells or cellular changes to perform: memory and other important brain functions. I also came up with an interesting idea to explain why people go into comas. I could be completely off. It's just an uneducated guess. The cause of comas could be that the brain circuit that controls attention has been damaged. The attention part of the brain probably drives everything by deciding what circuits to activate and why! Without that circuit creating activity, the brain's neurons have no reason to fire normally and the brain's normal activity does not occur. Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Compressed Cross-Indexed Concepts
This seems to be an overly simplistic view of AGI from a mathematician. It's kind of funny how people over emphasize what they know or depend on their current expertise too much when trying to solve new problems. I don't think it makes sense to apply sanitized and formal mathematical solutions to AGI. What reason do we have to believe that the problems we face when developing AGI are solvable by such formal representations? What reason do we have to think we can represent the problems as an instance of such mathematical problems? We have to start with the specific problems we are trying to solve, analyze what it takes to solve them, and then look for and design a solution. Starting with the solution and trying to hack the problem to fit it is not going to work for AGI, in my opinion. I could be wrong, but I would need some evidence to think otherwise. Dave On Wed, Aug 11, 2010 at 10:39 AM, Jim Bromer jimbro...@gmail.com wrote: You probably could show that a sophisticated mathematical structure would produce a scalable AGI program if is true, using contemporary mathematical models to simulate it. However, if scalability was completely dependent on some as yet undiscovered mathemagical principle, then you couldn't. For example, I think polynomial time SAT would solve a lot of problems with contemporary AGI. So I believe this could be demonstrated on a simulation. That means, that I could demonstrate effective AGI that works so long as the SAT problems are easily solved. If the program reported that a complicated logical problem could not be solved, the user could provide his insight into the problem at those times to help with the problem. This would not work exactly as hoped, but by working from there, I believe that I would be able to determine better ways to develop such a program so it would work better - if my conjecture about the potential efficacy of polynomial time SAT for AGI was true. Jim Bromer On Mon, Aug 9, 2010 at 6:11 PM, Jim Bromer jimbro...@gmail.com wrote: On Mon, Aug 9, 2010 at 4:57 PM, John G. Rose johnr...@polyplexic.comwrote: -Original Message- From: Jim Bromer [mailto:jimbro...@gmail.com] how would these diverse examples be woven into highly compressed and heavily cross-indexed pieces of knowledge that could be accessed quickly and reliably, especially for the most common examples that the person is familiar with. This is a big part of it and for me the most exciting. And I don't think that this subsystem would take up millions of lines of code either. It's just that it is a *very* sophisticated and dynamic mathematical structure IMO. John Well, if it was a mathematical structure then we could start developing prototypes using familiar mathematical structures. I think the structure has to involve more ideological relationships than mathematical. For instance you can apply a idea to your own thinking in a such a way that you are capable of (gradually) changing how you think about something. This means that an idea can be a compression of some greater change in your own programming. While the idea in this example would be associated with a fairly strong notion of meaning, since you cannot accurately understand the full consequences of the change it would be somewhat vague at first. (It could be a very precise idea capable of having strong effect, but the details of those effects would not be known until the change had progressed.) I think the more important question is how does a general concept be interpreted across a range of different kinds of ideas. Actually this is not so difficult, but what I am getting at is how are sophisticated conceptual interrelations integrated and resolved? Jim *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Compressed Cross-Indexed Concepts
Jim, Fair enough. My apologies then. I just often see your posts on SAT or other very formal math problems and got the impression that you thought this was at the core of AGI's problems and that pursuing a fast solution to NP-complete problems is the best way to solve it. At least, that was my impression. So, my thought was that such formal methods don't seem to be a complete solution at all and other factors, such as uncertainty, could make such formal solutions ineffective or unusable. Which is why I said it's important to analyze the requirements of the problem and then apply a solution. Dave On Wed, Aug 11, 2010 at 1:02 PM, Jim Bromer jimbro...@gmail.com wrote: David, I am not a mathematician although I do a lot of computer-related mathematical work of course. My remark was directed toward John who had suggested that he thought that there is some sophisticated mathematical sub system that would (using my words here) provide such a substantial benefit to AGI that its lack may be at the core of the contemporary problem. I was saying that unless this required mathemagic then a scalable AGI system demonstrating how effective this kind of mathematical advancement could probably be simulated using contemporary mathematics. This is not the same as saying that AGI is solvable by sanitized formal representations any more than saying that your message is a sanitized formal statement because it was dependent on a lot of computer mathematics in order to send it. In other words I was challenging John at that point to provide some kind of evidence for his view. I then went on to say, that for example, I think that fast SAT solutions would make scalable AGI possible (that is, scalable up to a point that is way beyond where we are now), and therefore I believe that I could create a simulation of an AGI program to demonstrate what I am talking about. (A simulation is not the same as the actual thing.) I didn't say, nor did I imply, that the mathematics would be all there is to it. I have spent a long time thinking about the problems of applying formal and informal systems to 'real world' (or other world) problems and the application of methods is a major part of my AGI theories. I don't expect you to know all of my views on the subject but I hope you will keep this in mind for future discussions. Jim Bromer On Wed, Aug 11, 2010 at 10:53 AM, David Jones davidher...@gmail.comwrote: This seems to be an overly simplistic view of AGI from a mathematician. It's kind of funny how people over emphasize what they know or depend on their current expertise too much when trying to solve new problems. I don't think it makes sense to apply sanitized and formal mathematical solutions to AGI. What reason do we have to believe that the problems we face when developing AGI are solvable by such formal representations? What reason do we have to think we can represent the problems as an instance of such mathematical problems? We have to start with the specific problems we are trying to solve, analyze what it takes to solve them, and then look for and design a solution. Starting with the solution and trying to hack the problem to fit it is not going to work for AGI, in my opinion. I could be wrong, but I would need some evidence to think otherwise. Dave On Wed, Aug 11, 2010 at 10:39 AM, Jim Bromer jimbro...@gmail.comwrote: You probably could show that a sophisticated mathematical structure would produce a scalable AGI program if is true, using contemporary mathematical models to simulate it. However, if scalability was completely dependent on some as yet undiscovered mathemagical principle, then you couldn't. For example, I think polynomial time SAT would solve a lot of problems with contemporary AGI. So I believe this could be demonstrated on a simulation. That means, that I could demonstrate effective AGI that works so long as the SAT problems are easily solved. If the program reported that a complicated logical problem could not be solved, the user could provide his insight into the problem at those times to help with the problem. This would not work exactly as hoped, but by working from there, I believe that I would be able to determine better ways to develop such a program so it would work better - if my conjecture about the potential efficacy of polynomial time SAT for AGI was true. Jim Bromer On Mon, Aug 9, 2010 at 6:11 PM, Jim Bromer jimbro...@gmail.com wrote: On Mon, Aug 9, 2010 at 4:57 PM, John G. Rose johnr...@polyplexic.comwrote: -Original Message- From: Jim Bromer [mailto:jimbro...@gmail.com] how would these diverse examples be woven into highly compressed and heavily cross-indexed pieces of knowledge that could be accessed quickly and reliably, especially for the most common examples that the person is familiar with. This is a big part of it and for me the most exciting. And I don't think
Re: [agi] Re: Compressed Cross-Indexed Concepts
Slightly off the topic of your last email. But, all this discussion has made me realize how to phrase something... That is that solving AGI requires understand the constraints that problems impose on a solution. So, it's sort of a unbelievably complex constraint satisfaction problem. What we've been talking about is how we come up with solutions to these problems when we sometimes aren't actually trying to solve any of the real problems. As I've been trying to articulate lately is that in order to satisfy the constraints of the problems AGI imposes, we must really understand the problems we want to solve and how they can be solved(their constraints). I think that most of us do not do this because the problem is so complex, that we refuse to attempt to understand all of its constraints. Instead we focus on something very small and manageable with fewer constraints. But, that's what creates narrow AI, because the constraints you have developed the solution for only apply to a narrow set of problems. Once you try to apply it to a different problem that imposes new, incompatible constraints, the solution fails. So, lately I've been pushing for people to truly analyze the problems involved in AGI, step by step to understand what the constraints are. I think this is the only way we will develop a solution that is guaranteed to work without wasting undo time in trial and error. I don't think trial and error approaches will work. We must know what the constraints are, instead of guessing at what solutions might approximate the constraints. I think the problem space is too large to guess. Of course, I think acquisition of knowledge through automated means is the first step in understanding these constraints. But, unfortunately, few agree with me. Dave On Wed, Aug 11, 2010 at 3:44 PM, Jim Bromer jimbro...@gmail.com wrote: I've made two ultra-brilliant statements in the past few days. One is that a concept can simultaneously be both precise and vague. And the other is that without judgement even opinions are impossible. (Ok, those two statements may not be ultra-brilliant but they are brilliant right? Ok, maybe not truly brilliant, but highly insightful and perspicuously intelligent... Or at least interesting to the cognoscenti maybe?.. Well, they were interesting to me at least.) Ok, these two interesting-to-me comments made by me are interesting because they suggest that we do not know how to program a computer even to create opinions. Or if we do, there is a big untapped difference between those programs that show nascent judgement (perhaps only at levels relative to the domain of their capabilities) and those that don't. This is AGI programmer's utopia. (Or at least my utopia). Because I need to find something that is simple enough for me to start with and which can lend itself to develop and test theories of AGI judgement and scalability. By allowing an AGI program to participate more in the selection of its own primitive 'interests' we will be able to interact with it, both as programmer and as user, to guide it toward selecting those interests which we can understand and seem interesting to us. By creating an AGI program that has a faculty for primitive judgement (as we might envision such an ability), and then testing the capabilities in areas where the program seems to work more effectively, we might be better able to develop more powerful AGI theories that show greater scalability, so long as we are able to understand what interests the program is pursuing. Jim Bromer On Wed, Aug 11, 2010 at 1:40 PM, Jim Bromer jimbro...@gmail.com wrote: On Wed, Aug 11, 2010 at 10:53 AM, David Jones davidher...@gmail.comwrote: I don't think it makes sense to apply sanitized and formal mathematical solutions to AGI. What reason do we have to believe that the problems we face when developing AGI are solvable by such formal representations? What reason do we have to think we can represent the problems as an instance of such mathematical problems? We have to start with the specific problems we are trying to solve, analyze what it takes to solve them, and then look for and design a solution. Starting with the solution and trying to hack the problem to fit it is not going to work for AGI, in my opinion. I could be wrong, but I would need some evidence to think otherwise. I agree that disassociated theories have not proved to be very successful at AGI, but then again what has? I would use a mathematical method that gave me the number or percentage of True cases that satisfy a propositional formula as a way to check the internal logic of different combinations of logic-based conjectures. Since methods that can do this with logical variables for any logical system that goes (a little) past 32 variables are feasible the potential of this method should be easy to check (although it would hit a rather low ceiling of scalability). So I do think that logic
Re: [agi] Nao Nao
Way too pessimistic in my opinion. On Mon, Aug 9, 2010 at 7:06 PM, John G. Rose johnr...@polyplexic.comwrote: Aww, so cute. I wonder if it has a Wi-Fi connection, DHCP's an IP address, and relays sensory information back to the main servers with all the other Nao's all collecting personal data in a massive multi-agent geo-distributed robo-network. So cuddly! And I wonder if it receives and executes commands, commands that come in over the network from whatever interested corporation or government pays the most for access. Such a sweet little friendly Nao. Everyone should get one :) John *From:* Mike Tintner [mailto:tint...@blueyonder.co.uk] An unusually sophisticated ( somewhat expensive) promotional robot vid: http://www.telegraph.co.uk/technology/news/7934318/Nao-the-robot-that-expresses-and-detects-emotions.html *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/| Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Anyone going to the Singularity Summit?
Steve, Capable and effective AI systems would be very helpful at every step of the research process. Basic research is a major area I think that AGI will be applied to. In fact, that's exactly where I plan to apply it first. Dave On Tue, Aug 10, 2010 at 7:25 AM, Steve Richfield steve.richfi...@gmail.comwrote: Ben, On Mon, Aug 9, 2010 at 1:07 PM, Ben Goertzel b...@goertzel.org wrote: I'm speaking there, on Ai applied to life extension; and participating in a panel discussion on narrow vs. general AI... Having some interest, expertise, and experience in both areas, I find it hard to imagine much interplay at all. The present challenge is wrapped up in a lack of basic information, resulting from insufficient funds to do the needed experiments. Extrapolations have already gone WAY beyond the data, and new methods to push extrapolations even further wouldn't be worth nearly as much as just a little more hard data. Just look at Aubrey's long list of aging mechanisms. We don't now even know which predominate, or which cause others. Further, there are new candidates arising every year, e.g. Burzynski's theory that most aging is secondary to methylation of DNA receptor sites, or my theory that Aubrey's entire list could be explained by people dropping their body temperatures later in life. There are LOTS of other theories, and without experimental results, there is absolutely no way, AI or not, to sort the wheat from the chaff. Note that one of the front runners, the cosmic ray theory, could easily be tested by simply raising some mice in deep tunnels. This is high-school level stuff, yet with NO significant funding for aging research, it remains undone. Note my prior posting explaining my inability even to find a source of used mice for kids to use in high-school anti-aging experiments, all while university labs are now killing their vast numbers of such mice. So long as things remain THIS broken, anything that isn't part of the solution simply becomes a part of the very big problem, AIs included. The best that an AI could seemingly do is to pronounce Fund and facilitate basic aging research and then suspend execution pending an interrupt indicating that the needed experiments have been done. Could you provide some hint as to where you are going with this? Steve *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Anyone going to the Singularity Summit?
The think the biggest thing to remember here is that general AI could be applied to many different problems in parallel by many different people. They would help with many aspects of the problem solving process, not just a single one and certainly not just applied to a single experiment/study. I'm confident that Ben is aware of this On Tue, Aug 10, 2010 at 1:43 PM, Bob Mottram fuzz...@gmail.com wrote: On 10 August 2010 16:44, Ben Goertzel b...@goertzel.org wrote: I'm writing an article on the topic for H+ Magazine, which will appear in the next couple weeks ... I'll post a link to it when it appears I'm not advocating applying AI in the absence of new experiments of course. I've been working closely with Genescient, applying AI tech to analyze the genomics of their long-lived superflies, so part of my message is about the virtuous cycle achievable via synergizing AI data analysis with carefully-designed experimental evolution of model organisms... Probably if I was going to apply AI in a medical context I'd prioritize those conditions which are both common and either fatal or have a severe impact on quality of life. Also worthwhile would be using AI to try to discover drugs which have an equivalent effect to existing known ones but can be manufactured at a significantly lower cost, such that they are brought within the means of a larger fraction of the population. Investigating aging is perfectly legitimate, but if you're trying to maximize your personal utility I'd regard it as a low priority compared to other more urgent medical issues which cause premature deaths. Also in the endeavor to extend life we need not focus entirely upon medical aspects. The organizational problems of delivering known medications on a large scale is also a problem which AI could perhaps be used to optimize. The way in which things like this are currently organized seems to be based upon some combination of tradition and intuitive hunches, so there may be low hanging fruit to be obtained here. For example, if an epidemic breaks out, why should you vaccinate first? If you have access to a social graph (from Facebook, or wherever) it's probably possible to calculate an optimal strategy. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Anyone going to the Singularity Summit?
Bob, their are serious issues with such a suggestion. The biggest issue, is that there is a good chance it wouldn't work because diseases, including the common cold, have incubation times. So, you may not have any symptoms at all, yet you can pass it on to other people. And even if we did know who was sick, are you really going to stay home for 2 weeks every time you get sick? If I were an employer, I would rather have you come to work when you feel up to it. Another point I've given to germaphobes is that let's say you are successful at avoiding as many possible germs as possible and avoid getting sick as much as possible. That means that you are likely not immune to some common colds and such that you should be. So, when you are old and less capable, your immune system will not be able to fight off the infection and you will die an early death. Dave On Tue, Aug 10, 2010 at 1:51 PM, Bob Mottram fuzz...@gmail.com wrote: On 10 August 2010 18:43, Bob Mottram fuzz...@gmail.com wrote: here. For example, if an epidemic breaks out, why should you vaccinate first? That should have been who rather than why :-) Just thinking a little further, in hand waving mode, If something like the common cold were added as a status within social networks, and everyone was on the network it might even be possible to eliminate this disease simply by getting people to avoid those who are known to have it for a certain period of time - a sort of internet enabled smart avoidance strategy. This wouldn't be a cure, but it could severely hamper the disease transmission mechanism, perhaps even to the extent of driving it to extinction. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: RE: [agi] How To Create General AI Draft2
I agree John that this is a useful exercise. This would be a good discussion if mike would ever admit that I might be right and he might be wrong. I'm not sure that will ever happen though. :) First he says I can't define a pattern that works. Then, when I do, he says the pattern is no good because it isn't physical. Lol. If he would ever admit that I might have gotten it right, the discussion would be a good one. Instead, he hugs his preconceived notions no matter how good my arguments are and finds yet another reason, any reason will do, to say I'm still wrong. On Aug 9, 2010 2:18 AM, John G. Rose johnr...@polyplexic.com wrote: Actually this is quite critical. Defining a chair - which would agree with each instance of a chair in the supplied image - is the way a chair should be defined and is the way the mind processes it. It can be defined mathematically in many ways. There is a particular one I would go for though... John *From:* Mike Tintner [mailto:tint...@blueyonder.co.uk] *Sent:* Sunday, August 08, 2010 7:28 AM To: agi Subject: Re: [agi] How To Create General AI Draft2 You're waffling. You say there's a pattern for chair - DRAW IT. Attached should help you. Analyse the chairs given in terms of basic visual units. Or show how any basic units can be applied to them. Draw one or two. You haven't identified any basic visual units - you don't have any. Do you? Yes/no. No. That's not funny, that's a waste.. And woolly and imprecise through and through. *From:* David Jones davidher...@gmail.com *Sent:* Sunday, August 08, 2010 1:59 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] How To Create General AI Draft2 Mike, We've argued about this over and over and over. I don't want to repeat previous arguments to you. You have no proof that the world cannot be broken down into simpler concepts and components. The only proof you attempt to propose are your example problems that *you* don't understand how to solve. Just because *you* cannot solve them, doesn't mean they cannot be solved at all using a certain methodology. So, who is really making wild assumptions? The mere fact that you can refer to a chair means that it is a recognizable pattern. LOL. That fact that you don't realize this is quite funny. Dave On Sun, Aug 8, 2010 at 8:23 AM, Mike Tintner tint...@blueyonder.co.uk wrote: Dave:No... it is equivalent to saying that the whole world can be modeled as if everything was made up of matter And matter is... ? Huh? You clearly don't realise that your thinking is seriously woolly - and you will pay a heavy price in lost time. What are your basic world/visual-world analytic units wh. you are claiming to exist? You thought - perhaps think still - that *concepts* wh. are pretty fundamental intellectual units of analysis at a certain level, could be expressed as, or indeed, were patterns. IOW there's a fundamental pattern for chair or table. Absolute nonsense. And a radical failure to understand the basic nature of concepts which is that they are *freeform* schemas, incapable of being expressed either as patterns or programs. You had merely assumed that concepts could be expressed as patterns,but had never seriously, visually analysed it. Similarly you are merely assuming that the world can be analysed into some kind of visual units - but you haven't actually done the analysis, have you? You don't have any of these basic units to hand, do you? If you do, I suggest, reply instantly, naming a few. You won't be able to do it. They don't exist. Your whole approach to AGI is based on variations of what we can call fundamental analysis - and it's wrong. God/Evolution hasn't built the world with any kind of geometric, or other consistent, bricks. He/It is a freeform designer. You have to start thinking outside the box/brick/fundamental unit. *From:* David Jones davidher...@gmail.com *Sent:* Sunday, August 08, 2010 5:12 AM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] How To Create General AI Draft2 Mike, I took your comments into consideration and have been updating my paper to make sure these problems are addressed. See more comments below. On Fri, Aug 6, 2010 at 8:15 PM, Mike Tintner tint...@blueyonder.co.uk wrote: 1) You don't define the difference between narrow AI and AGI - or make clear why your approach is one and not the other I removed this because my audience is for AI researchers... this is AGI 101. I think it's clear that my design defines general as being able to handle the vast majority of things we want the AI to handle without requiring a change in design. 2) Learning about the world won't cut it - vast nos. of progs. claim they can learn about the world - what's the difference between narrow AI and AGI learning? The difference is in what you can or can't learn about and what tasks you can or can't perform. If the AI is able to receive input about anything it needs to know about in the same formats
Re: [agi] How To Create General AI Draft2
You see. This is precisely why I don't want to argue with Mike anymore. it must be a physical pattern. LOL. Who ever said that patterns must be physical? This is exactly why you can't see my point of view. You impose unnecessary restrictions on any possible solution when there really are no such restrictions. Dave On Mon, Aug 9, 2010 at 7:27 AM, Mike Tintner tint...@blueyonder.co.ukwrote: John:It can be defined mathematically in many ways Try it - crude drawings/jottings/diagrams totally acceptable. See my set of fotos to Dave. (And yes, you're right this is of extreme importance. And no. Dave, there are no such things as non-physical patterns). --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How To Create General AI Draft2
I already stated these. read previous emails. On Mon, Aug 9, 2010 at 8:48 AM, Mike Tintner tint...@blueyonder.co.ukwrote: PS Examples of nonphysical patterns AND how they are applicable to visual AGI.? *From:* David Jones davidher...@gmail.com *Sent:* Monday, August 09, 2010 1:34 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] How To Create General AI Draft2 You see. This is precisely why I don't want to argue with Mike anymore. it must be a physical pattern. LOL. Who ever said that patterns must be physical? This is exactly why you can't see my point of view. You impose unnecessary restrictions on any possible solution when there really are no such restrictions. Dave On Mon, Aug 9, 2010 at 7:27 AM, Mike Tintner tint...@blueyonder.co.ukwrote: John:It can be defined mathematically in many ways Try it - crude drawings/jottings/diagrams totally acceptable. See my set of fotos to Dave. (And yes, you're right this is of extreme importance. And no. Dave, there are no such things as non-physical patterns). *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How To Create General AI Draft2
Mike, Quoting a previous email: QUOTE In fact, the chair patterns you refer to are not strictly physical patterns. The pattern is based on how the objects can be used, what their intended uses probably are, and what most common effective uses are. So, chairs are objects that are used to sit on. You can identify objects whose most likely use is for sitting based on experience. END QUOTE Even refrigerators can be chairs. If a fridge is in the woods and you're out there camping, you can sit on it. I could say sit on that fridge couch over there. The fact that multiple people can sit on it, makes it possible to call it a couch. But, it's odd to call it a chair, because it's a fridge. So, when the object has a more common effective use, as I stated above, it is usually referred to by that use. If something is most likely used for sitting by a single person, then it is a chair. If its most common best use is something else, like cooling food, you would call it a fridge. So, maybe the pattern would be, if it has some features like a chair, like possible arm rests, a soft bottom, cushions, legs, a back rest, etc. and you can't see it being used as anything else, then maybe it's a chair. If someone sits on it, it certainly is a chair, if you find it by searching for chairs, its likely a chair. etc. You see, chairs are not simply recognized by their physical structure. There are multiple ways you can recognize it and it is certainly important to know that it doesn't seem useful for another task. The idea that chairs cannot be recognized because they come in all shapes, sizes and structures is just wrong. Dave On Mon, Aug 9, 2010 at 8:47 AM, Mike Tintner tint...@blueyonder.co.ukwrote: Examples of nonphysical patterns? *From:* David Jones davidher...@gmail.com *Sent:* Monday, August 09, 2010 1:34 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] How To Create General AI Draft2 You see. This is precisely why I don't want to argue with Mike anymore. it must be a physical pattern. LOL. Who ever said that patterns must be physical? This is exactly why you can't see my point of view. You impose unnecessary restrictions on any possible solution when there really are no such restrictions. Dave On Mon, Aug 9, 2010 at 7:27 AM, Mike Tintner tint...@blueyonder.co.ukwrote: John:It can be defined mathematically in many ways Try it - crude drawings/jottings/diagrams totally acceptable. See my set of fotos to Dave. (And yes, you're right this is of extreme importance. And no. Dave, there are no such things as non-physical patterns). *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How To Create General AI Draft2
, But what's not so obvious - although undeniable - is how stretchable and fluid that line must be in order to recognize diverse objects - as diverse as one octopus, one cactus, one mountain. See foto below. The brain can stretch a line outwards to encompass any form of object in the universe - or conversely, squeeze/stretch any object inwards to form a 1. All those objects in the foto can be squeezed/stretched into that one on the top left. Now is anyone here going to have the gall to tell me that process of object recognition is mathematical? But just as strings are - or could be - central to matter and physics; so are fluid schemas central to intelligence - and especially to concepts. **Correction - a blind idiot *could* see - by touch - that the diverse forms of one octopus/flower etc could not be reduced to a line by any mathematical process. P.S. When I say that maths cannot deal with fluid schemas and object recognition, one should perhaps amend that - it may be that no existing form of maths. wh. deals entirely in set forms and patterns can, but that a creative version of maths, dealing in free forms and patchworks, could. P.P.S. String - the concept - itself involves an extremely fluid schema - is a variation, in fact, of the schema of one/1 - and must embrace many diverse forms that strings may be shaped into. *From:* David Jones davidher...@gmail.com *Sent:* Monday, August 09, 2010 2:13 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] How To Create General AI Draft2 Mike, Quoting a previous email: QUOTE In fact, the chair patterns you refer to are not strictly physical patterns. The pattern is based on how the objects can be used, what their intended uses probably are, and what most common effective uses are. So, chairs are objects that are used to sit on. You can identify objects whose most likely use is for sitting based on experience. END QUOTE Even refrigerators can be chairs. If a fridge is in the woods and you're out there camping, you can sit on it. I could say sit on that fridge couch over there. The fact that multiple people can sit on it, makes it possible to call it a couch. But, it's odd to call it a chair, because it's a fridge. So, when the object has a more common effective use, as I stated above, it is usually referred to by that use. If something is most likely used for sitting by a single person, then it is a chair. If its most common best use is something else, like cooling food, you would call it a fridge. So, maybe the pattern would be, if it has some features like a chair, like possible arm rests, a soft bottom, cushions, legs, a back rest, etc. and you can't see it being used as anything else, then maybe it's a chair. If someone sits on it, it certainly is a chair, if you find it by searching for chairs, its likely a chair. etc. You see, chairs are not simply recognized by their physical structure. There are multiple ways you can recognize it and it is certainly important to know that it doesn't seem useful for another task. The idea that chairs cannot be recognized because they come in all shapes, sizes and structures is just wrong. Dave On Mon, Aug 9, 2010 at 8:47 AM, Mike Tintner tint...@blueyonder.co.ukwrote: Examples of nonphysical patterns? *From:* David Jones davidher...@gmail.com *Sent:* Monday, August 09, 2010 1:34 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] How To Create General AI Draft2 You see. This is precisely why I don't want to argue with Mike anymore. it must be a physical pattern. LOL. Who ever said that patterns must be physical? This is exactly why you can't see my point of view. You impose unnecessary restrictions on any possible solution when there really are no such restrictions. Dave On Mon, Aug 9, 2010 at 7:27 AM, Mike Tintner tint...@blueyonder.co.ukwrote: John:It can be defined mathematically in many ways Try it - crude drawings/jottings/diagrams totally acceptable. See my set of fotos to Dave. (And yes, you're right this is of extreme importance. And no. Dave, there are no such things as non-physical patterns). *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com/ *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com/ *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com/ *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com
Re: [agi] How To Create General AI Draft2
Thanks Ben, I think the biggest difference with the way I approach it is to be deliberate in how the system solves specific kinds of problems. I haven't gone into that in detail yet though. For example, Itamar seems to want to give the AI the basic building blocks that make up spaciotemporal dependencies as a sort of bag of features and just let a neural-net-like structure find the patterns. If that is not accurate, please correct me. I am very skeptical of such approaches because there is no guarantee at all that the system will properly represent the relationships and structure of the data. It seems just hopeful to me that such a system would get it right out of the vast number of possible results it could accidental arrive at. The human visual system doesn't evolve like that on the fly. This can be proven by the fact that we all see the same visual illusions. We all exhibit the same visual limitations in the same way. There is much evidence that the system doesn't evolve accidentally. It has a limited set of rules it uses to learn from perceptual data. I think a more deliberate approach would be more effective because we can understand why it does what it does, how it does it, and why its not working if it doesn't work. With such deliberate approaches, it is much more clear how to proceed and to reuse knowledge in many complementary ways. This is what I meant by emergence. I propose a more deliberate approach that knows exactly why problems can be solved a certain way and how the system is likely to solve them. I'm suggesting to represent the spaciotemporal relationships deliberately and explicitly. Then we can construct general algorithms to solve problems explicitly, yet generally. Regarding computer vision not being that important... Don't you think that because knowledge is so essential and manual input is inneffective, perception-based acquisition of knowledge is a very serious barrier to AGI? It seems to me that the solutions to AGI problems being constructed are not using knowledge gained from simulated perception effectively. OpenCog's natural language processing for example, seems to use very very little knowledge that would be gathered from visual perception. As far as I remember, it mostly uses things that are learned from other sources. To me, it doesn't make sense to spend so much time debugging and developing such solutions, when a better and more general approach to language understanding would use a lot of knowledge. Those are the sorts of things I feel are new to this approach. Thanks Again, Dave PS: I'm planning to go to the Singularity Summit :) Last minute. Hope to see you there. On Mon, Aug 9, 2010 at 10:01 AM, Ben Goertzel b...@goertzel.org wrote: Hi David, I read the essay I think it summarizes well some of the key issues involving the bridge between perception and cognition, and the hierarchical decomposition of natural concepts I find the ideas very harmonious with those of Jeff Hawkins, Itamar Arel, and other researchers focused on hierarchical deep learning approaches to vision with longer-term AGI ambitions I'm not sure there are any dramatic new ideas in the essay. Do you think there are? My own view is that these ideas are basically right, but handle only a modest percentage of what's needed to make a human-level, vaguely human-like AGI I.e. I don't agree that solving vision and the vision-cognition bridge is *such* a huge part of AGI, though it's certainly a nontrivial percentage... -- Ben G On Fri, Aug 6, 2010 at 4:44 PM, David Jones davidher...@gmail.com wrote: Hey Guys, I've been working on writing out my approach to create general AI to share and debate it with others in the field. I've attached my second draft of it in PDF format, if you guys are at all interested. It's still a work in progress and hasn't been fully edited. Please feel free to comment, positively or negatively, if you have a chance to read any of it. I'll be adding to and editing it over the next few days. I'll try to reply more professionally than I have been lately :) Sorry :S Cheers, Dave *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC CTO, Genescient Corp Vice Chairman, Humanity+ Advisor, Singularity University and Singularity Institute External Research Professor, Xiamen University, China b...@goertzel.org I admit that two times two makes four is an excellent thing, but if we are to give everything its due, two times two makes five is sometimes a very charming thing too. -- Fyodor Dostoevsky *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com
Re: [agi] How To Create General AI Draft2
Ben, Comments below. On Mon, Aug 9, 2010 at 12:00 PM, Ben Goertzel b...@goertzel.org wrote: The human visual system doesn't evolve like that on the fly. This can be proven by the fact that we all see the same visual illusions. We all exhibit the same visual limitations in the same way. There is much evidence that the system doesn't evolve accidentally. It has a limited set of rules it uses to learn from perceptual data. That is not a proof, of course. It could be that given a general architecture, and inputs with certain statistical properties, the same internal structures inevitably self-organize You're right, I should organize details and evidence that the human brain has a lot of its processing algorithms built in. Another example of this innate ability to process inputs the right way is the fact that many language acquisition researchers believe that children have a built-in hypothesis space that they use when learning language (see generativism at http://en.wikipedia.org/wiki/Language_acquisition). It is likely not enough to just give it all the data it needs and let it guess till it fines a good answer. The hypothesis space is likely too large. So I'm curious -- what are the specific pattern-recognition modules that you will put into your system, and how will you arrange them hierarchically? Well, the first pattern-recognition modules are the ones for inferring scene and object structures and properties from visual/lidar data. I can't really be specific because The next set of pattern-recognition modules would be for inferring relationships such as object whole-to-part relationships and their other behavioral relationships. Basically, algorithms for inferring a sparse or dense models of objects. Again, it is quite hard to be specific about algorithms. There is a lot of detailed analysis that I have yet to do for each type of problem and how the whole is broken down into these types of relationships. Again, as you can see, I think the problem can be broken down into generic components that can be reasoned about. As for hierarchical design... I haven't decided yet. It really depends on the purpose of the hierarchy and its function. That's why in the paper I stress function before design. -- how will you handle feedback connections (top-down) among the modules? That's a very good question. I haven't decided yet really because I haven't fully worked out all the pieces of the design and how they must interact to solve problems. I'd need to analyze specific requirements and what problems such feedback is required to solve. I guess one example of feedback might be the interpretation of ambiguous visual input, such as single images from a less than ideal camera and scene setup. Such problems require feedback from knowledge. I see this as a separate visual processing system from the visual learning system that I mentioned in the paper. This is because the system I designed is for learning from less ambiguous input. Once it has gained sufficient knowledge this way, more ambiguous input would be possible to process and understand with confidence. So, clearly much still has to be worked out about the design. But, my working assumption is that these things can be broken down analytically and solved. The alternative is to just hope that a similar-to-the-brain model is going to work. I just don't think we can reasonably hope that such a model will work, be effective and be efficient. I think it is just too hard to guess at the right structure that will solve the problems without actually showing how it solves all the problems we want to apply it to. * I really think it is very important for the functional requirements to create the design.* Regardless of the approach, we need to understand why the solutions we create solve the problems we want to solve. And if we can't show that they do solve them or how they solve them, then the odds are against us that they will work. That's my opinion. If one could show how deep learning models, for example, really do solve all the problems we want to solve, then I would be willing to use them. I just don't see it though. I doesn't seem that the solution was generated by the problem. It seems more that the solution was generated based on its similarity to the brain. I just can't accept the risk that such approaches won't work. Since I don't think reverse engineering the brain makes sense either. My only alternative to those two approaches seems to be the one I'm taking. Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] Anyone going to the Singularity Summit?
I've decided to go. I was wondering if anyone else here is. Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How To Create General AI Draft2
Mike, We've argued about this over and over and over. I don't want to repeat previous arguments to you. You have no proof that the world cannot be broken down into simpler concepts and components. The only proof you attempt to propose are your example problems that *you* don't understand how to solve. Just because *you* cannot solve them, doesn't mean they cannot be solved at all using a certain methodology. So, who is really making wild assumptions? The mere fact that you can refer to a chair means that it is a recognizable pattern. LOL. That fact that you don't realize this is quite funny. Dave On Sun, Aug 8, 2010 at 8:23 AM, Mike Tintner tint...@blueyonder.co.ukwrote: Dave:No... it is equivalent to saying that the whole world can be modeled as if everything was made up of matter And matter is... ? Huh? You clearly don't realise that your thinking is seriously woolly - and you will pay a heavy price in lost time. What are your basic world/visual-world analytic units wh. you are claiming to exist? You thought - perhaps think still - that *concepts* wh. are pretty fundamental intellectual units of analysis at a certain level, could be expressed as, or indeed, were patterns. IOW there's a fundamental pattern for chair or table. Absolute nonsense. And a radical failure to understand the basic nature of concepts which is that they are *freeform* schemas, incapable of being expressed either as patterns or programs. You had merely assumed that concepts could be expressed as patterns,but had never seriously, visually analysed it. Similarly you are merely assuming that the world can be analysed into some kind of visual units - but you haven't actually done the analysis, have you? You don't have any of these basic units to hand, do you? If you do, I suggest, reply instantly, naming a few. You won't be able to do it. They don't exist. Your whole approach to AGI is based on variations of what we can call fundamental analysis - and it's wrong. God/Evolution hasn't built the world with any kind of geometric, or other consistent, bricks. He/It is a freeform designer. You have to start thinking outside the box/brick/fundamental unit. *From:* David Jones davidher...@gmail.com *Sent:* Sunday, August 08, 2010 5:12 AM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] How To Create General AI Draft2 Mike, I took your comments into consideration and have been updating my paper to make sure these problems are addressed. See more comments below. On Fri, Aug 6, 2010 at 8:15 PM, Mike Tintner tint...@blueyonder.co.ukwrote: 1) You don't define the difference between narrow AI and AGI - or make clear why your approach is one and not the other I removed this because my audience is for AI researchers... this is AGI 101. I think it's clear that my design defines general as being able to handle the vast majority of things we want the AI to handle without requiring a change in design. 2) Learning about the world won't cut it - vast nos. of progs. claim they can learn about the world - what's the difference between narrow AI and AGI learning? The difference is in what you can or can't learn about and what tasks you can or can't perform. If the AI is able to receive input about anything it needs to know about in the same formats that it knows how to understand and analyze, it can reason about anything it needs to. 3) Breaking things down into generic components allows us to learn about and handle the vast majority of things we want to learn about. This is what makes it general! Wild assumption, unproven or at all demonstrated and untrue. You are only right that I haven't demonstrated it. I will address this in the next paper and continue adding details over the next few drafts. As a simple argument against your counter argument... If that were true that we could not understand the world using a limited set of rules or concepts, how is it that a human baby, with a design that is predetermined to interact with the world a certain way by its DNA, is able to deal with unforeseen things that were not preprogrammed? That’s right, the baby was born with a set of rules that robustly allows it to deal with the unforeseen. It has a limited set of rules used to learn. That is equivalent to a limited set of “concepts” (i.e. rules) that would allow a computer to deal with the unforeseen. Interesting philosophically because it implicitly underlies AGI-ers' fantasies of take-off. You can compare it to the idea that all science can be reduced to physics. If it could, then an AGI could indeed take-off. But it's demonstrably not so. No... it is equivalent to saying that the whole world can be modeled as if everything was made up of matter. Oh, I forgot, that is the case :) It is a limited set of concepts, yet it can create everything we know. You don't seem to understand that the problem of AGI is to deal with the NEW - the unfamiliar
Re: [agi] How To Create General AI Draft2
:) what you don't realize is that patterns don't have to be strictly limited to the actual physical structure. In fact, the chair patterns you refer to are not strictly physical patterns. The pattern is based on how the objects can be used, what their intended uses probably are, and what most common effective uses are. So, chairs are objects that are used to sit on. You can identify objects whose most likely use is for sitting based on experience. If you think this is not a sufficient refutation of your argument, then please don't argue with me regarding it anymore. I know your opinion and respectfully disagree. If you don't accept my counter argument, there is no point to continuing this back and forth ad finitum. Dave On Aug 8, 2010 9:29 AM, Mike Tintner tint...@blueyonder.co.uk wrote: You're waffling. You say there's a pattern for chair - DRAW IT. Attached should help you. Analyse the chairs given in terms of basic visual units. Or show how any basic units can be applied to them. Draw one or two. You haven't identified any basic visual units - you don't have any. Do you? Yes/no. No. That's not funny, that's a waste.. And woolly and imprecise through and through. *From:* David Jones davidher...@gmail.com *Sent:* Sunday, August 08, 2010 1:59 PM To: agi Subject: Re: [agi] How To Create General AI Draft2 Mike, We've argued about this over and over and over. I don't want to repeat previous arguments to you. You have no proof that the world cannot be broken down into simpler concepts and components. The only proof you attempt to propose are your example problems that *you* don't understand how to solve. Just because *you* cannot solve them, doesn't mean they cannot be solved at all using a certain methodology. So, who is really making wild assumptions? The mere fact that you can refer to a chair means that it is a recognizable pattern. LOL. That fact that you don't realize this is quite funny. Dave On Sun, Aug 8, 2010 at 8:23 AM, Mike Tintner tint...@blueyonder.co.ukwrote: Dave:No... it is equivalent to saying that the whole world can be modeled as if everything was made up of matter And matter is... ? Huh? You clearly don't realise that your thinking is seriously woolly - and you will pay a heavy price in lost time. What are your basic world/visual-world analytic units wh. you are claiming to exist? You thought - perhaps think still - that *concepts* wh. are pretty fundamental intellectual units of analysis at a certain level, could be expressed as, or indeed, were patterns. IOW there's a fundamental pattern for chair or table. Absolute nonsense. And a radical failure to understand the basic nature of concepts which is that they are *freeform* schemas, incapable of being expressed either as patterns or programs. You had merely assumed that concepts could be expressed as patterns,but had never seriously, visually analysed it. Similarly you are merely assuming that the world can be analysed into some kind of visual units - but you haven't actually done the analysis, have you? You don't have any of these basic units to hand, do you? If you do, I suggest, reply instantly, naming a few. You won't be able to do it. They don't exist. Your whole approach to AGI is based on variations of what we can call fundamental analysis - and it's wrong. God/Evolution hasn't built the world with any kind of geometric, or other consistent, bricks. He/It is a freeform designer. You have to start thinking outside the box/brick/fundamental unit. *From:* David Jones davidher...@gmail.com *Sent:* Sunday, August 08, 2010 5:12 AM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] How To Create General AI Draft2 Mike, I took your comments into consideration and have been updating my paper to make sure these problems are addressed. See more comments below. On Fri, Aug 6, 2010 at 8:15 PM, Mike Tintner tint...@blueyonder.co.ukwrote: 1) You don't define the difference between narrow AI and AGI - or make clear why your approach is one and not the other I removed this because my audience is for AI researchers... this is AGI 101. I think it's clear that my design defines general as being able to handle the vast majority of things we want the AI to handle without requiring a change in design. 2) Learning about the world won't cut it - vast nos. of progs. claim they can learn about the world - what's the difference between narrow AI and AGI learning? The difference is in what you can or can't learn about and what tasks you can or can't perform. If the AI is able to receive input about anything it needs to know about in the same formats that it knows how to understand and analyze, it can reason about anything it needs to. 3) Breaking things down into generic components allows us to learn about and handle the vast majority of things we want to learn about. This is what makes it general! Wild assumption, unproven or at all demonstrated and untrue
Re: [agi] Help requested: Making a list of (non-robotic) AGI low hanging fruit apps
Hey Ben, Faster, cheaper, and more robust 3D modeling for the movie industry. The modeling allows different sources of video content to be extracted from scenes, manipulated and mixed with others. The movie industry has the money and motivation to extract data from images. Making it easier, more robust and cheaper could drive innovation and progress. Why is it AGI-related? Because AGI requires knowledge. Knowledge can be extracted from facts about the world. Facts can be extracted from images in a general way using a limited set of algorithms and concepts. Some say that computer vision is AI-complete and requires knowledge to do. But, I have to disagree. Given sufficient data and good images from multiple cameras or devices, unambiguous data can extract very accurate 3D models. If this was AI-completed and required knowledge, that would not be possible. Dave On Sat, Aug 7, 2010 at 9:10 PM, Ben Goertzel b...@goertzel.org wrote: Hi, A fellow AGI researcher sent me this request, so I figured I'd throw it out to you guys I'm putting together an AGI pitch for investors and thinking of low hanging fruit applications to argue for. I'm intentionally not involving any mechanics (robots, moving parts, etc.). I'm focusing on voice (i.e. conversational agents) and perhaps vision-based systems. Hellen Keller AGI, if you will :) Along those lines, I'd like any ideas you may have that would fall under this description. I need to substantiate the case for such AGI technology by making an argument for high-value apps. All ideas are welcome. All serious responses will be appreciated!! Also, I would be grateful if we could keep this thread closely focused on direct answers to this question, rather than digressive discussions on Helen Keller, the nature of AGI, the definition of AGI versus narrow AI, the achievability or unachievability of AGI, etc. etc. If you think the question is bad or meaningless or unclear or whatever, that's fine, but please start a new thread with a different subject line to make your point. If the discussion is useful, my intention is to mine the answers into a compact list to convey to him Thanks! Ben G --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How To Create General AI Draft2
Abram, Thanks for the comments. I think probability is just one way to deal with uncertainty. Defeasible reasoning is another. Non-monotonic logic of various implementations. I often think that probability is the wrong way to do some things regarding AGI design. Maybe things can't be known with super high confidence, but we still want as high confidence as reasonably possible. Once we have that, we just have to have working assumptions and working hypotheses. From there we need the ability to update beliefs if we can find a reason to think the beliefs are wrong... Dave On Fri, Aug 6, 2010 at 9:48 PM, Abram Demski abramdem...@gmail.com wrote: On Fri, Aug 6, 2010 at 8:22 PM, Abram Demski abramdem...@gmail.comwrote: (Without this sort of generality, your approach seems restricted to gathering knowledge about whatever events unfold in front of a limited quantity of high-quality camera systems which you set up. To be honest, the usefulness of that sort of knowledge is not obvious.) On second thought, this statement was a bit naive. You obviously intend the camera systems to be connected to robots or other systems which perform actual tasks in the world, providing a great variety of information including feedback from success/failure of actions to achieve results. What is unrealistic to me is not that this information could be useful, but that this level of real-world intelligence could be achieved with the super-high confidence bounds you are imagining. What I think is that probabilistic reasoning is needed. Once we have the object/location/texture information with those confidence bounds (which I do see as possible), gaining the sort of knowledge Cyc set out to contain seems inherently statistical. --Abram On Fri, Aug 6, 2010 at 4:44 PM, David Jones davidher...@gmail.comwrote: Hey Guys, I've been working on writing out my approach to create general AI to share and debate it with others in the field. I've attached my second draft of it in PDF format, if you guys are at all interested. It's still a work in progress and hasn't been fully edited. Please feel free to comment, positively or negatively, if you have a chance to read any of it. I'll be adding to and editing it over the next few days. I'll try to reply more professionally than I have been lately :) Sorry :S Cheers, Dave *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- Abram Demski http://lo-tho.blogspot.com/ http://groups.google.com/group/one-logic -- Abram Demski http://lo-tho.blogspot.com/ http://groups.google.com/group/one-logic *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How To Create General AI Draft2
Mike, I took your comments into consideration and have been updating my paper to make sure these problems are addressed. See more comments below. On Fri, Aug 6, 2010 at 8:15 PM, Mike Tintner tint...@blueyonder.co.ukwrote: 1) You don't define the difference between narrow AI and AGI - or make clear why your approach is one and not the other I removed this because my audience is for AI researchers... this is AGI 101. I think it's clear that my design defines general as being able to handle the vast majority of things we want the AI to handle without requiring a change in design. 2) Learning about the world won't cut it - vast nos. of progs. claim they can learn about the world - what's the difference between narrow AI and AGI learning? The difference is in what you can or can't learn about and what tasks you can or can't perform. If the AI is able to receive input about anything it needs to know about in the same formats that it knows how to understand and analyze, it can reason about anything it needs to. 3) Breaking things down into generic components allows us to learn about and handle the vast majority of things we want to learn about. This is what makes it general! Wild assumption, unproven or at all demonstrated and untrue. You are only right that I haven't demonstrated it. I will address this in the next paper and continue adding details over the next few drafts. As a simple argument against your counter argument... If that were true that we could not understand the world using a limited set of rules or concepts, how is it that a human baby, with a design that is predetermined to interact with the world a certain way by its DNA, is able to deal with unforeseen things that were not preprogrammed? That’s right, the baby was born with a set of rules that robustly allows it to deal with the unforeseen. It has a limited set of rules used to learn. That is equivalent to a limited set of “concepts” (i.e. rules) that would allow a computer to deal with the unforeseen. Interesting philosophically because it implicitly underlies AGI-ers' fantasies of take-off. You can compare it to the idea that all science can be reduced to physics. If it could, then an AGI could indeed take-off. But it's demonstrably not so. No... it is equivalent to saying that the whole world can be modeled as if everything was made up of matter. Oh, I forgot, that is the case :) It is a limited set of concepts, yet it can create everything we know. You don't seem to understand that the problem of AGI is to deal with the NEW - the unfamiliar, that wh. cannot be broken down into familiar categories, - and then find ways of dealing with it ad hoc. You don't seem to understand that even the things you think cannot be broken down, can be. Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Computer Vision not as hard as I thought!
On Fri, Aug 6, 2010 at 7:37 PM, Jim Bromer jimbro...@gmail.com wrote: On Wed, Aug 4, 2010 at 9:27 AM, David Jones davidher...@gmail.com wrote: *So, why computer vision? Why can't we just enter knowledge manually? * a) The knowledge we require for AI to do what we want is vast and complex and we can prove that it is completely ineffective to enter the knowledge we need manually. b) Computer vision is the most effective means of gathering facts about the world. Knowledge and experience can be gained from analysis of these facts. c) Language is not learned through passive observation. The associations that words have to the environment and our common sense knowledge of the environment/world are absolutely essential to language learning, understanding and disambiguation. When visual information is available, children use visual cues from their parents and from the objects they are interacting with to figure out word-environment associations. If visual info is not available, touch is essential to replace the visual cues. Touch can provide much of the same info as vision, but it is not as effective because not everything is in reach and it provides less information than vision. There is some very good documentation out there on how children learn language that supports this. One example is How Children Learn Language by William O'grady. d) The real world cannot be predicted blindly. It is absolutely essential to be able to directly observe it and receive feedback. e) Manual entry of knowledge, even if possible, would be extremely slow and would be a very serious bottleneck(it already is). This is a major reason we want AI... to increase our man power and remove man-power related bottlenecks. Discovering a way to get a computer program to interpret a human language is a difficult problem. The feeling that an AI program might be able to attain a higher level of intelligence if only it could examine data from a variety of different kinds of sensory input modalities it is not new. It has been tried and tried during the past 35 years. But there is no experimental data (that I have heard of) that suggests that this method is the only way anyone will achieve intelligence. if only it could examine data from a variety of different kinds of sensory input modalities That statement suggests that such different kinds of input have no meaningful relationship to the problem at hand. I'm not talking about different kinds of input. I'm talking about explicitly and deliberately extracting facts about the environment from sensory perception, specifically remote perception or visual perception. The input modalities are not what is important. It is the facts that you can extract from computer vision that are useful in understanding what is out there in the world, what relationships and associations exist, and how is language associated with the environment. It is well documented that children learn language by interacting with adults around them and using cues from them to learn how the words they speak are associated with what is going on. It is not hard to support the claim that extensive knowledge about the world is important for understanding and interpreting human language. Nor is it hard to support the idea that such knowledge can be gained from computer vision. I have tried to explain that I believe the problem is twofold. First of all, there have been quite a few AI programs that worked real well as long as the problem was simple enough. This suggests that the complexity of what is trying to be understood is a critical factor. This in turn suggests that using different input modalities, would not -in itself- make AI possible. Your conclusion isn't supported by your arguments. I'm not even saying it makes AI possible. I'm saying that a system can make reasonable inferences and come to reasonable conclusions with sufficient knowledge. Without sufficient knowledge, there is reason to believe that it is significantly harder and often impossible to come to correct conclusions. Therefore, gaining knowledge about how things are related is not just helpful in making correct inferences, it is required. So, different input modalities which can give you facts about the world, which in turn would give you knowledge about the world, do make correct reasoning possible, when it otherwise would not be possible. You see, it has nothing to do with the source of the info or whether it is more info or not. It has everything to do with the relationships that information have. Just calling them different input modalities is not correct. Secondly, there is a problem of getting the computer to accurately model that which it can know in such a way that it could be effectively utilized for higher degrees of complexity. This is an engineering problem, not necessarily a problem that can't be solved. When we get
Re: [agi] Computer Vision not as hard as I thought!
they are interacting with to figure out word-environment associations. If visual info is not available, touch is essential to replace the visual cues. Touch can provide much of the same info as vision, but it is not as effective because not everything is in reach and it provides less information than vision. There is some very good documentation out there on how children learn language that supports this. One example is How Children Learn Language by William O'grady. d) The real world cannot be predicted blindly. It is absolutely essential to be able to directly observe it and receive feedback. e) Manual entry of knowledge, even if possible, would be extremely slow and would be a very serious bottleneck(it already is). This is a major reason we want AI... to increase our man power and remove man-power related bottlenecks. I could argue the above pieces separately. But, since the email is already long, I'll leave at that for now. If you want to explore any of them further, I can delve more into them. On Wed, Aug 4, 2010 at 9:10 AM, Jim Bromer jimbro...@gmail.com wrote: On Tue, Aug 3, 2010 at 11:52 AM, David Jones davidher...@gmail.comwrote: I've suddenly realized that computer vision of real images is very much solvable and that it is now just a matter of engineering... I've also realized that I don't actually have to implement it, which is what is most difficult because even if you know a solution to part of the problem has certain properties and issues, implementing it takes a lot of time. Whereas I can just assume I have a less than perfect solution with the properties I predict from other experiments. Then I can solve the problem without actually implementing every last detail... *First*, existing methods find observations that are likely true by themselves. They find data patterns that are very unlikely to occur by coincidence, such as many features moving together over several frames of a video and over a statistically significant distance. They use thresholds to ensure that the observed changes are likely transformations of the original property observed or to ensure the statistical significance of an observation. These are highly likely true observations and not coincidences or noise. -- Just looking at these statements, I can find three significant errors. (I do agree with some of what you said, like the significance of finding observations that are likely true in themselves.) When the camera moves (in a rotation or pan) many features will appear 'to move together over a statistically significant distance'. One might suppose that the animal can sense the movement of his own eyes but then again, he can rotate his head and use his vision to gauge the rotation of his head. So right away there is some kind of serious error in your statement. It might be resolvable, it is just that your statement does not really do the resolution. I do believe that computer vision is possible with contemporary computers but you are also saying that while you can't get your algorithms to work the way you had hoped, it doesn't really matter because you can figure it all out without the work of implementation. My point of view is that these represent major errors in reasoning. I hope to get back to actual visual processing experiments again. Although I don't think that computer vision is necessary for AGI, I do think that there is so much to be learned from experimenting with computer vision that it is a serious mistake not to take advantage of opportunity. Jim Bromer On Tue, Aug 3, 2010 at 11:52 AM, David Jones davidher...@gmail.comwrote: I've suddenly realized that computer vision of real images is very much solvable and that it is now just a matter of engineering. I was so stuck before because you can't make the simple assumptions in screenshot computer vision that you can in real computer vision. This makes experience probably necessary to effectively learn from screenshots. Objects in real images to not change drastically in appearance, position or other dimensions in unpredictable ways. The reason I came to the conclusion that it's a lot easier than I thought is that I found a way to describe why existing solutions work, how they work and how to come up with even better solutions. I've also realized that I don't actually have to implement it, which is what is most difficult because even if you know a solution to part of the problem has certain properties and issues, implementing it takes a lot of time. Whereas I can just assume I have a less than perfect solution with the properties I predict from other experiments. Then I can solve the problem without actually implementing every last detail. *First*, existing methods find observations that are likely true by themselves. They find data patterns that are very unlikely to occur by coincidence, such as many features moving together over several
Re: [agi] Computer Vision not as hard as I thought!
Steve, I wouldn't say that's an accurate description of what I wrote. What a wrote was a way to think about how to solve computer vision. My approach to artificial intelligence is a Neat approach. See http://en.wikipedia.org/wiki/Neats_vs._scruffies The paper you attached is a Scruffy approach. Neat approaches are characterized by deliberate algorithms that are analogous to the problem and can sometimes be shown to be provably correct. An example of a Neat approach is the use of features in the paper I mentioned. One can describe why the features are calculated and manipulated the way they are. An example of a scruffies approach would be neural nets, where you don't know the rules by which it comes up with an answer and such approaches are not very scalable. Neural nets require manually created training data and the knowledge generated is not in a form that can be used for other tasks. The knowledge isn't portable. I also wouldn't say I switched from absolute values to rates of change. That's not really at all what I'm saying here. Dave On Wed, Aug 4, 2010 at 2:32 PM, Steve Richfield steve.richfi...@gmail.comwrote: David, It appears that you may have reinvented the wheel. See the attached article. There is LOTS of evidence, along with some good math, suggesting that our brains work on rates of change rather than absolute values. Then, temporal learning, which is otherwise very difficult, falls out as the easiest of things to do. In effect, your proposal shifts from absolute values to rates of change. Steve === On Tue, Aug 3, 2010 at 8:52 AM, David Jones davidher...@gmail.com wrote: I've suddenly realized that computer vision of real images is very much solvable and that it is now just a matter of engineering. I was so stuck before because you can't make the simple assumptions in screenshot computer vision that you can in real computer vision. This makes experience probably necessary to effectively learn from screenshots. Objects in real images to not change drastically in appearance, position or other dimensions in unpredictable ways. The reason I came to the conclusion that it's a lot easier than I thought is that I found a way to describe why existing solutions work, how they work and how to come up with even better solutions. I've also realized that I don't actually have to implement it, which is what is most difficult because even if you know a solution to part of the problem has certain properties and issues, implementing it takes a lot of time. Whereas I can just assume I have a less than perfect solution with the properties I predict from other experiments. Then I can solve the problem without actually implementing every last detail. *First*, existing methods find observations that are likely true by themselves. They find data patterns that are very unlikely to occur by coincidence, such as many features moving together over several frames of a video and over a statistically significant distance. They use thresholds to ensure that the observed changes are likely transformations of the original property observed or to ensure the statistical significance of an observation. These are highly likely true observations and not coincidences or noise. *Second*, they make sure that the other possible explanations of the observations are very unlikely. This is usually done using a threshold, and a second difference threshold from the first match to the second match. This makes sure that second best matches are much farther away than the best match. This is important because it's not enough to find a very likely match if there are 1000 very likely matches. You have to be able to show that the other matches are very unlikely, otherwise the specific match you pick may be just a tiny bit better than the others, and the confidence of that match would be very low. So, my initial design plans are as follows. Note: I will probably not actually implement the system because the engineering part dominates the time. I'd rather convert real videos to pseudo test cases or simulation test cases and then write a psuedo design and algorithm that can solve it. This would show that it works without actually spending the time needed to implement it. It's more important for me to prove it works and show what it can do than to actually do it. If I can prove it, there will be sufficient motivation for others to do it with more resources and man power than I have at my disposal. *My Design* *First, we use high speed cameras and lidar systems to gather sufficient data with very low uncertainty because the changes possible between data points can be assumed to be very low, allowing our thresholds to be much smaller, which eliminates many possible errors and ambiguities. *Second*, *we have to gain experience from high confidence observations. These are gathered as follows: 1) Describe allowable transformations(thresholds) and what they mean
Re: [agi] Computer Vision not as hard as I thought!
Steve, Sorry if I misunderstood your approach. I do not really understand how it would work though because it is not clear how you go from inputs to output goals. It likely will still have many of the same problems as other neural networks including 1) poor knowledge portability 2) difficult to extend, augment or understand how it works 3) requires manually created training data, which is a major problem. 4) is designed with biological hardware in mind, not necessarily existing hardware and software. These are my main reasons, at least that I can remember, that I avoid biologically inspired methods. It's not to say that they are wrong. But they don't meet my requirements. It is also very unclear how to implement the system and make it work. My approach is very deliberate, so the steps required to make it work are pretty clear to me. It is not that your approach is bad. It is just different and I really prefer methods that are not biologically inspired, but are designed specifically with goals and requirements in mind as the most important design motivator. Dave On Wed, Aug 4, 2010 at 3:54 PM, Steve Richfield steve.richfi...@gmail.comwrote: David, You are correct in that I keep bad company. My approach to NNs is VERY different than other people's approaches. I insist on reasonable math being performed on quantities that I understand, which sets me apart from just about everyone else. Your neat approach isn't all that neat, and is arguably scruffier than mine. At least I have SOME math to back up my approach. Further, note that we are self-organizing systems, and that this process is poorly understood. I am NOT particularly interest in people-programmed systems because of their very fundamental limitations. Yes, self-organization is messy, but it fits the neat definition better than it meets the scruffy definition. Scruffy has more to do with people-programmed ad hoc approaches (like most of AGI), which I agree are a waste of time. Steve On Wed, Aug 4, 2010 at 12:43 PM, David Jones davidher...@gmail.comwrote: Steve, I wouldn't say that's an accurate description of what I wrote. What a wrote was a way to think about how to solve computer vision. My approach to artificial intelligence is a Neat approach. See http://en.wikipedia.org/wiki/Neats_vs._scruffies The paper you attached is a Scruffy approach. Neat approaches are characterized by deliberate algorithms that are analogous to the problem and can sometimes be shown to be provably correct. An example of a Neat approach is the use of features in the paper I mentioned. One can describe why the features are calculated and manipulated the way they are. An example of a scruffies approach would be neural nets, where you don't know the rules by which it comes up with an answer and such approaches are not very scalable. Neural nets require manually created training data and the knowledge generated is not in a form that can be used for other tasks. The knowledge isn't portable. I also wouldn't say I switched from absolute values to rates of change. That's not really at all what I'm saying here. Dave On Wed, Aug 4, 2010 at 2:32 PM, Steve Richfield steve.richfi...@gmail.com wrote: David, It appears that you may have reinvented the wheel. See the attached article. There is LOTS of evidence, along with some good math, suggesting that our brains work on rates of change rather than absolute values. Then, temporal learning, which is otherwise very difficult, falls out as the easiest of things to do. In effect, your proposal shifts from absolute values to rates of change. Steve === On Tue, Aug 3, 2010 at 8:52 AM, David Jones davidher...@gmail.comwrote: I've suddenly realized that computer vision of real images is very much solvable and that it is now just a matter of engineering. I was so stuck before because you can't make the simple assumptions in screenshot computer vision that you can in real computer vision. This makes experience probably necessary to effectively learn from screenshots. Objects in real images to not change drastically in appearance, position or other dimensions in unpredictable ways. The reason I came to the conclusion that it's a lot easier than I thought is that I found a way to describe why existing solutions work, how they work and how to come up with even better solutions. I've also realized that I don't actually have to implement it, which is what is most difficult because even if you know a solution to part of the problem has certain properties and issues, implementing it takes a lot of time. Whereas I can just assume I have a less than perfect solution with the properties I predict from other experiments. Then I can solve the problem without actually implementing every last detail. *First*, existing methods find observations that are likely true by themselves. They find data patterns that are very unlikely to occur
Re: [agi] Computer Vision not as hard as I thought!
Steve, I replace your need for math with my need to understand what the system is doing and why it is doing it. It's basically the same thing. But you are approaching it at an extremely low level. It doesn't seem to me that you are clear on how this math makes the system work the way we want it to work. So, make the math as perfect as you like, if you don't understand why you need the math and how it makes the system do what you want, then it's not going to do you any good. Understanding what you are trying to accomplish and how you want the system to work comes first, not math. If your neural net doesn't require training data, I don't understand how it works or why you expect it to do what you want it to do if it is self organized. How do you tell it how to process inputs correctly? What guides the processing and analysis? Dave On Wed, Aug 4, 2010 at 4:33 PM, Steve Richfield steve.richfi...@gmail.comwrote: David On Wed, Aug 4, 2010 at 1:16 PM, David Jones davidher...@gmail.com wrote: 3) requires manually created training data, which is a major problem. Where did this come from. Certainly, people are ill equipped to create dP/dt type data. These would have to come from sensors. 4) is designed with biological hardware in mind, not necessarily existing hardware and software. The biology is just good to help the math over some humps. So far, I have not been able to identify ANY neuronal characteristic that hasn't been refined to near-perfection, once the true functionality was fully understood. Anyway, with the math, you can build a system anyway you want. Without the math, you are just wasting your time and electricity. The math comes first, and all other things follow. Steve === These are my main reasons, at least that I can remember, that I avoid biologically inspired methods. It's not to say that they are wrong. But they don't meet my requirements. It is also very unclear how to implement the system and make it work. My approach is very deliberate, so the steps required to make it work are pretty clear to me. It is not that your approach is bad. It is just different and I really prefer methods that are not biologically inspired, but are designed specifically with goals and requirements in mind as the most important design motivator. Dave On Wed, Aug 4, 2010 at 3:54 PM, Steve Richfield steve.richfi...@gmail.com wrote: David, You are correct in that I keep bad company. My approach to NNs is VERY different than other people's approaches. I insist on reasonable math being performed on quantities that I understand, which sets me apart from just about everyone else. Your neat approach isn't all that neat, and is arguably scruffier than mine. At least I have SOME math to back up my approach. Further, note that we are self-organizing systems, and that this process is poorly understood. I am NOT particularly interest in people-programmed systems because of their very fundamental limitations. Yes, self-organization is messy, but it fits the neat definition better than it meets the scruffy definition. Scruffy has more to do with people-programmed ad hoc approaches (like most of AGI), which I agree are a waste of time. Steve On Wed, Aug 4, 2010 at 12:43 PM, David Jones davidher...@gmail.comwrote: Steve, I wouldn't say that's an accurate description of what I wrote. What a wrote was a way to think about how to solve computer vision. My approach to artificial intelligence is a Neat approach. See http://en.wikipedia.org/wiki/Neats_vs._scruffies The paper you attached is a Scruffy approach. Neat approaches are characterized by deliberate algorithms that are analogous to the problem and can sometimes be shown to be provably correct. An example of a Neat approach is the use of features in the paper I mentioned. One can describe why the features are calculated and manipulated the way they are. An example of a scruffies approach would be neural nets, where you don't know the rules by which it comes up with an answer and such approaches are not very scalable. Neural nets require manually created training data and the knowledge generated is not in a form that can be used for other tasks. The knowledge isn't portable. I also wouldn't say I switched from absolute values to rates of change. That's not really at all what I'm saying here. Dave On Wed, Aug 4, 2010 at 2:32 PM, Steve Richfield steve.richfi...@gmail.com wrote: David, It appears that you may have reinvented the wheel. See the attached article. There is LOTS of evidence, along with some good math, suggesting that our brains work on rates of change rather than absolute values. Then, temporal learning, which is otherwise very difficult, falls out as the easiest of things to do. In effect, your proposal shifts from absolute values to rates of change. Steve === On Tue, Aug 3, 2010 at 8:52 AM, David Jones davidher
Re: [agi] Computer Vision not as hard as I thought!
On Wed, Aug 4, 2010 at 6:17 PM, Steve Richfield steve.richfi...@gmail.comwrote: David, On Wed, Aug 4, 2010 at 1:45 PM, David Jones davidher...@gmail.com wrote: Understanding what you are trying to accomplish and how you want the system to work comes first, not math. It's all the same. First comes the qualitative, then comes the quantitative. If your neural net doesn't require training data, Sure it needs training data -real-world interactive sensory input training data, rather than static manually prepared training data. You design is not described well enough or succinctly enough for me to comment on then. I don't understand how it works or why you expect it to do what you want it to do if it is self organized. How do you tell it how to process inputs correctly? What guides the processing and analysis? Bingo - you have just hit on THE great challenge in AI/AGI., and the source of much past debate. Some believe in maximizing the information content of the output. Some believe in other figures of merit, e.g. success in interacting with a test environment, success in forming a layered structure, etc. This particular sub-field is still WIDE open and waiting for some good answers. Note that this same problem presents itself, regardless of approach, e.g. AGI. Ah, but I think that this problem is much more solvable and better defined with a more deliberate approach that does not depend on emergence. Emergence is wishful thinking. I hope you do not include such wishful thinking in your design :) Once the AI has the tools and knowledge needed to solve a problem, which I expect to get from computer vision, then it can reason about user stated goals (in natural language) and we can work on how the goal pursuit part works. Much work has already been done on planning and execution. But, all that work was done with insufficient knowledge on narrow problems. All the research needs to be re-evaluated and studied with sufficient knowledge about the world. It changes everything. This is another mile marker on my roadmap to general AI. Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Walker Lake
How about you go to war yourself or send your children. I'd rather send a robot. It's safer for both the soldier and the people on the ground because you don't have to shoot first, ask questions later. And you're right, we shouldn't monitor anyone. We should just allow terrorists to talk openly to plot attacks on us. After all, I'd rather have my privacy than my life. dumb. On Mon, Aug 2, 2010 at 10:40 AM, Steve Richfield steve.richfi...@gmail.comwrote: Sometime when you are flying between the northwest US to/from Las Vegas, look out your window as you fly over Walker Lake in eastern Nevada. At the south end you will see a system of roads leading to tiny buildings, all surrounded by military security. From what I have been able to figure out, you will find the U.S. arsenal of chemical and biological weapons housed there. No, we are not now making these weapons, but neither are we disposing of them. Similarly, there has been discussion of developing advanced military technology using AGI and other computer-related methods. I believe that these efforts are fundamentally anti-democratic, as they allow a small number of people to control a large number of people. Gone are the days when people voted with their swords. We now have the best government that money can buy monitoring our every email, including this one, to identify anyone resisting such efforts. 1984 has truly arrived. This can only lead to a horrible end to freedom, with AGIs doing their part and more. Like chemical and biological weapons, unmanned and automated weapons should be BANNED. Unfortunately, doing so would provide a window of opportunity for others to deploy them. However, if we make these and stick them in yet another building at the south end of Walker Lake, we would be ready in case other nations deploy such weapons. How about an international ban on the deployment of all unmanned and automated weapons? The U.S. won't now even agree to ban land mines. At least this would restore SOME relationship between popular support and military might. Doesn't it sound ethical to insist that a human being decide when to end another human being's life? Doesn't it sound fair to require the decision maker to be in harm's way, especially when the person being killed is in or around their own home? Doesn't it sound unethical to add to the present situation? When deployed on a large scale, aren't these WMDs? Steve *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Shhh!
Abram Wrote: I take this as evidence that there is a very strong mental landscape... if you go in a particular direction there is a natural series of landmarks, including both great ideas and pitfalls that everyone runs into. (Different people take different amounts of time to climb out of the pitfalls, though. Some may keep looking for gold at a dead end for a long time.) That is a very nice description of AI research and the pitfalls we come across in our quest. :) Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Clues to the Mind: Learning Ability
:) Intelligence isn't limited to higher cognitive functions. One could say a virus is intelligent or alive because it can replicate itself. Intelligence is not just one function or ability, it can be many different things. But mostly, for us, it comes down to what the system can accomplish for us. As for the turing test, it is basically worthless in my opinion. PS: you probably should post these video posts to a single thread... Dave On Wed, Jul 28, 2010 at 12:39 AM, deepakjnath deepakjn...@gmail.com wrote: http://www.facebook.com/video/video.php?v=287151911466 See how the parrot can learn so much! Does that mean that the parrot does intelligence. Will this parrot pass the turing test? There must be a learning center in the brain which is much lower than the higher cognitive fucntions like imagination and thoughts. cheers, Deepak *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Huge Progress on the Core of AGI
Sure. Thanks Arthur. On Sun, Jul 25, 2010 at 10:42 AM, A. T. Murray menti...@scn.org wrote: David Jones wrote: Arthur, Thanks. I appreciate that. I would be happy to aggregate some of those things. I am sometimes not good at maintaining the website because I get bored of maintaining or updating it very quickly :) Dave On Sat, Jul 24, 2010 at 10:02 AM, A. T. Murray menti...@scn.org wrote: The Web site of David Jones at http://practicalai.org is quite impressive to me as a kindred spirit building AGI. (Just today I have been coding MindForth AGI :-) For his Practical AI Challenge or similar ventures, I would hope that David Jones is open to the idea of aggregating or archiving representative AI samples from such sources as - TexAI; - OpenCog; - Mentifex AI; - etc.; so that visitors to PracticalAI may gain an overview of what is happening in our field. Arthur -- http://www.scn.org/~mentifex/AiMind.htmlhttp://www.scn.org/%7Ementifex/AiMind.html http://www.scn.org/~mentifex/mindforth.txthttp://www.scn.org/%7Ementifex/mindforth.txt Just today, a few minutes ago, I updated the mindforth.txt AI souce code listed above. In the PracticalAi aggregates, you might consider listing Mentifex AI with copies of the above two AI source code pages, and with links to the original scn.org URL's, where visitors to PracticalAi could look for any more recent updates that you had not gotten around to transferring from scn.org to PracticalAi. In that way, theses releases of Mentifex free AI source code would have a more robust Web presence (SCN often goes down) and I could link to PracticalAi for the aggregates and other features of PracticalAI. Thanks. Arthur T. Murray --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How do we hear music
Deepak, I have some insight on this question. There was a study regarding change blindness. One of the study's famous experiments was having a person ask for directions on a college campus. Then in the middle of this, a door would pass between the person asking directions and the student giving directions. What they found is that many people didn't realize the person had changed. BUT, 100% of the people that did notice the change were the same age or younger than the person they were observing! So, they did another experiment to rule out the different possible explanations. They took young people and dressed them as construction workers. Then, they performed the experiment again with similar age groups. They found that the people that had noticed the change before no longer did! Why? Well, the evidence leads us to believe that people pay much closer attention to the details of people they consider to be similar to them. So, we notice fewer details when we are observing people of a group we consider our out-group. In other words, we don't think we belong to the same group as the person we are observing. That is why asians all look the same to you :) I think the purpose of this is analogous to attention. We only learn about things we consider important. Or we only pay attention to things we think are important. So, for whatever reason, we think that out-group people are not as important to us, and we don't need to spend our brain's resources on remembering details about them. Dave On Jul 26, 2010 2:58 PM, deepakjnath deepakjn...@gmail.com wrote: Mike, All chinese look the same for me. But for a chinese person they don't. Why is this? Is there another clue here? Thanks, Deepak On Mon, Jul 26, 2010 at 9:10 PM, Mike Tintner tint...@blueyonder.co.uk wrote: David, T... -- cheers, Deepak *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Clues to the Mind: What do you think is the reason for selective attention
I found proof of my interpretation in the following paper also. It concludes that we can only keep track of 3 or 4 objects in detail at a time.(something like that) http://www.pni.princeton.edu/conte/pdfs/project2/Proj2Pub8anne.pdf It says: For explicit visual working memory, object tokens are stored in a limited capacity, vulnerable store that maintains the bindings of features for just 2 to 4 objects. Attention is required to sustain the memories. Dave On Sun, Jul 25, 2010 at 1:00 AM, deepakjnath deepakjn...@gmail.com wrote: Thanks Dave, its very interesting. This gives us more clues in to how the brain compresses and uses the relevant information while neglecting the irrelevant information. But as Anast has demonstrated, the brain does need priming inorder to decide what is relevant and irrelevant. :) Cheers, Deepak On Sun, Jul 25, 2010 at 5:34 AM, David Jones davidher...@gmail.comwrote: I also wanted to say that it is agi related because this may be the way that the brain deals with ambiguity in the real world. It ignores many things if it can use expectations to constrain possibilities. It is an important way in which the brain tracks objects and identifies them without analyzing all of an objects features before matching over the whole image. On Jul 24, 2010 7:53 PM, David Jones davidher...@gmail.com wrote: Actually Deepak, this is AGI related. This week I finally found a cool body of research that I previously had no knowledge of. This research area is in psychology, which is probably why I missed it the first time. It has to do with human perception, object files, how we keep track of object, individuate them, match them (the correspondence problem), etc. And I found the perfect article just now for you Deepak: http://www.duke.edu/~mitroff/papers/SimonsMitroff_01.pdfhttp://www.duke.edu/%7Emitroff/papers/SimonsMitroff_01.pdf This article mentions why the brain does not notice things. And I just realized as I was reading it why we don't see the gorilla or other unexpected changes. The reason is this: We have a limited amount of processing power that we can apply to visual tracking and analysis. So, in attention demanding situations such as these, we assign our processing resources to only track the things we are interested in. In fact, we probably do this all the time, but it is only when we need a lot of attention to be applied to a few objects do we notice that we don't see some unexpected events. So, our brain knows where to expect the ball next and our visual processing is very busy tracking the ball and then seeing who is throwing it. As a result, it is unable to also process the movement of other objects. If the unexpected event is drastic enough, it will get our attention. But since some of the people are in black, our brain probably thinks it is just a person in black and doesn't consider it an event that is worthy of interrupting our intense tracking. Dave On Sat, Jul 24, 2010 at 4:58 PM, Anastasios Tsiolakidis sokratis.dk@ gmail.com wrote: On Sat,... *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- cheers, Deepak *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Huge Progress on the Core of AGI
lol. thanks Jim :) On Thu, Jul 22, 2010 at 10:08 PM, Jim Bromer jimbro...@gmail.com wrote: I have to say that I am proud of David Jone's efforts. He has really matured during these last few months. I'm kidding but I really do respect the fact that he is actively experimenting. I want to get back to work on my artificial imagination and image analysis programs - if I can ever figure out how to get the time. As I have read David's comments, I realize that we need to really leverage all sorts of cruddy data in order to make good agi. But since that kind of thing doesn't work with sparse knowledge, it seems that the only way it could work is with extensive knowledge about a wide range of situations, like the knowledge gained from a vast variety of experiences. This conjecture makes some sense because if wide ranging knowledge could be kept in superficial stores where it could be accessed quickly and economically, it could be used efficiently in (conceptual) model fitting. However, as knowledge becomes too extensive it might become too unwieldy to find what is needed for a particular situation. At this point indexing becomes necessary with cross-indexing references to different knowledge based on similarities and commonalities of employment. Here I am saying that relevant knowledge based on previous learning might not have to be totally relevant to a situation as long as it could be used to run during an ongoing situation. From this perspective then, knowledge from a wide variety of experiences should actually be composed of reactions on different conceptual levels. Then as a piece of knowledge is brought into play for an ongoing situation, those levels that seem best suited to deal with the situation could be promoted quickly as the situation unfolds, acting like an automated indexing system into other knowledge relevant to the situation. So the ongoing process of trying to determine what is going on and what actions should be made would simultaneously act like an automated index to find better knowledge more suited for the situation. Jim Bromer *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
Abram, I haven't found a method that I think works consistently yet. Basically I was trying methods like the one you suggested, which measures the number of correct predictions or expectations. But, then I ran into the problem of, what if the predictions you are counting are more of the same? Do you count them or not? For example, lets say that we see a piece of paper on a table in an image and we see that the paper looks different but moves with the table. So, we can hypothesize that they are attached. Now what if it is not a piece of paper, but a mural. Do you count every little piece of the mural that moves with the desk as a correct prediction? Is it a single prediction? What about the number of times they move together? It doesn't seem right to count each and every time, but we also have to be careful about coincidental movement together. Just because it seems to move together in one frame out of 1000 does not mean we should consider them temporarily attached. So, quantitatively defining simpler and predictive is quite challenging. I am honestly a bit stumped at how to do it at the moment. I will keep trying to find ways to at least approximate it, but I'm really not sure the best way. Of course, I haven't been working on this specific problem long, but other people have tried to quantify our explanatory methods in other areas and have also failed. I think part of the failure has to do with the fact that the things they want to explain using the same method should probably use different methods and should be more heuristic than mathematically precise. It's all quite overwhelming to analyze sometimes. I may have thought about fractions correct vs. incorrect also. The truth is, I haven't locked on and carefully analyzed the different ideas I've come up with because they all seem to have issues and it is difficult to analyze. I definitely need to try some out and just see what the results are and document them better. Dave On Thu, Jul 22, 2010 at 10:23 PM, Abram Demski abramdem...@gmail.comwrote: David, What are the different ways you are thinking of for measuring the predictiveness? I can think of a few different possibilities (such as measuring number incorrect vs measuring fraction incorrect, et cetera) but I'm wondering which variations you consider significant/troublesome/etc. --Abram On Thu, Jul 22, 2010 at 7:12 PM, David Jones davidher...@gmail.comwrote: It's certainly not as simple as you claim. First, assigning a probability is not always possible, nor is it easy. The factors in calculating that probability are unknown and are not the same for every instance. Since we do not know what combination of observations we will see, we cannot have a predefined set of probabilities, nor is it any easier to create a probability function that generates them for us. That is just as exactly what I meant by quantitatively define the predictiveness... it would be proportional to the probability. Second, if you can define a program ina way that is always simpler when it is smaller, then you can do the same thing without a program. I don't think it makes any sense to do it this way. It is not that simple. If it was, we could solve a large portion of agi easily. On Thu, Jul 22, 2010 at 3:16 PM, Matt Mahoney matmaho...@yahoo.com wrote: David Jones wrote: But, I am amazed at how difficult it is to quantitatively define more predictive and simpler for specific problems. It isn't hard. To measure predictiveness, you assign a probability to each possible outcome. If the actual outcome has probability p, you score a penalty of log(1/p) bits. To measure simplicity, use the compressed size of the code for your prediction algorithm. Then add the two scores together. That's how it is done in the Calgary challenge http://www.mailcom.com/challenge/ and in my own text compression benchmark. -- Matt Mahoney, matmaho...@yahoo.com *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Thu, July 22, 2010 3:11:46 PM *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI Because simpler is not better if it is less predictive. On Thu, Jul 22, 2010 at 1:21 PM, Abram Demski abramdem...@gmail.com wrote: Jim, Why more predictive *and then* simpler? --Abram On Thu, Jul 22, 2010 at 11:49 AM, David Jones davidher...@gmail.com wrote: An Update I think the following gets to the heart of general AI and what it takes to achieve it. It also provides us with evidence as to why general AI is so difficult. With this new knowledge in mind, I think I will be much more capable now of solving the problems and making it work. I've come to the conclusion lately that the best hypothesis is better because it is more predictive and then simpler than other hypotheses (in that order more predictive... then simpler). But, I am amazed at how difficult it is to quantitatively define more predictive and simpler for specific problems
Re: [agi] Re: Huge Progress on the Core of AGI
Abram, I should also mention that I ran into problems mainly because I was having a hard time deciding how to identify objects and determine what is really going on in a scene. This adds a whole other layer of complexity to hypotheses. It's not just about what is more predictive of the observations, it is about deciding what exactly you are observing in the first place. (although you might say its the same problem). I ran into this problem when my algorithm finds matches between items that are not the same. Or it may not find any matches between items that are the same, but have changed. So, how do you decide whether it is 1) the same object, 2) a different object or 3) the same object but it has changed. And how do you decide its relationship to something else... is it 1) dependently attached 2) semi-dependently attached(can move independently, but only in certain ways. Yet also moves dependently) 3) independent 4) sometimes dependent 5) was dependent, but no longer is, 6) was dependent on something else, but then was independent, but now is dependent on something new. These hypotheses are different ways of explaining the same observations, but are complicated by the fact that we aren't sure of the identity of the objects we are observing in the first place. Multiple hypotheses may fit the same observations, and its hard to decide why one is simpler or better than the other. The object you were observing at first may have disappeared. A new object may have appeared at the same time (this is why screenshots are a bit malicious). Or the object you were observing may have changed. In screenshots, sometimes the objects that you are trying to identify as different never appear at the same time because they always completely occlude each other. So, that can make it extremely difficult to decide whether they are the same object that has changed or different objects. Such ambiguities are common in AGI. It is unclear to me yet how to deal with them effectively, although I am continuing to work hard on it. I know its a bit of a mess, but I'm just trying to demonstrate the trouble I've run into. I hope that makes it more clear why I'm having so much trouble finding a way of determining what hypothesis is most predictive and simplest. Dave On Thu, Jul 22, 2010 at 10:23 PM, Abram Demski abramdem...@gmail.comwrote: David, What are the different ways you are thinking of for measuring the predictiveness? I can think of a few different possibilities (such as measuring number incorrect vs measuring fraction incorrect, et cetera) but I'm wondering which variations you consider significant/troublesome/etc. --Abram On Thu, Jul 22, 2010 at 7:12 PM, David Jones davidher...@gmail.comwrote: It's certainly not as simple as you claim. First, assigning a probability is not always possible, nor is it easy. The factors in calculating that probability are unknown and are not the same for every instance. Since we do not know what combination of observations we will see, we cannot have a predefined set of probabilities, nor is it any easier to create a probability function that generates them for us. That is just as exactly what I meant by quantitatively define the predictiveness... it would be proportional to the probability. Second, if you can define a program ina way that is always simpler when it is smaller, then you can do the same thing without a program. I don't think it makes any sense to do it this way. It is not that simple. If it was, we could solve a large portion of agi easily. On Thu, Jul 22, 2010 at 3:16 PM, Matt Mahoney matmaho...@yahoo.com wrote: David Jones wrote: But, I am amazed at how difficult it is to quantitatively define more predictive and simpler for specific problems. It isn't hard. To measure predictiveness, you assign a probability to each possible outcome. If the actual outcome has probability p, you score a penalty of log(1/p) bits. To measure simplicity, use the compressed size of the code for your prediction algorithm. Then add the two scores together. That's how it is done in the Calgary challenge http://www.mailcom.com/challenge/ and in my own text compression benchmark. -- Matt Mahoney, matmaho...@yahoo.com *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Thu, July 22, 2010 3:11:46 PM *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI Because simpler is not better if it is less predictive. On Thu, Jul 22, 2010 at 1:21 PM, Abram Demski abramdem...@gmail.com wrote: Jim, Why more predictive *and then* simpler? --Abram On Thu, Jul 22, 2010 at 11:49 AM, David Jones davidher...@gmail.com wrote: An Update I think the following gets to the heart of general AI and what it takes to achieve it. It also provides us with evidence as to why general AI is so difficult. With this new knowledge in mind, I think I will be much more capable now of solving the problems and making it work. I've
Re: [agi] Re: Huge Progress on the Core of AGI
Matt, Any method must deal with similar, if not the same, ambiguities. You need to show how neural nets solve this problem or how they solve agi goals while completely skipping the problem. Until then, it is not a successful method. Dave On Jul 24, 2010 7:18 PM, Matt Mahoney matmaho...@yahoo.com wrote: Mike Tintner wrote: Huh, Matt? What examples of this holistic scene analysis are there (or are y... I mean a neural model with increasingly complex features, as opposed to an algorithmic 3-D model (like video game graphics in reverse). Of course David rejects such ideas ( http://practicalai.org/Prize/Default.aspx ) even though the one proven working vision model uses it. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* Mike Tintner tint...@blueyonder.co.uk To: agi agi@v2.listbox.com *Sent:* Sat, July 24, 2010 6:16:07 PM Subject: Re: [agi] Re: Huge Progress on the Core of AGI Huh, Matt? What examples of this holistic scene analysis are there (or are you thinking about)? ... *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
Check this out! The title Space and time, not surface features, guide object persistence says it all. http://pbr.psychonomic-journals.org/content/14/6/1199.full.pdf Over just the last couple days I have begun to realize that they are so right. My idea before of using high frame rates is also spot on. The brain does not use features as much as we think. First we construct a model of the object, then we probably decide what features to index it with for future search. If we know that the object occurs at a particular location in space, then we can learn a great deal about it with very little ambiguity! Of course, processing images at all is hard, but that's besides the point... The point is that we can automatically learn about the world using high frame rates and a simple heuristic for identifying specific objects in a scene. Because we can reliably identify them, we can learn an extremely large amount in a very short period of time. We can learn about how lighting affects the colors, noise, size, shape, components, attachment relationships, etc. etc. So, it is very likely that screenshots are not simpler than real images! lol. The objects in real images usually don't change as much, as drastically or as quickly as the objects in screenshots. That means that we can use the simple heuristics of size, shape, location and continuity of time to match objects and learn about them. Dave On Sat, Jul 24, 2010 at 9:10 PM, Matt Mahoney matmaho...@yahoo.com wrote: Mike Tintner wrote: Which is? The one right behind your eyes. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* Mike Tintner tint...@blueyonder.co.uk *To:* agi agi@v2.listbox.com *Sent:* Sat, July 24, 2010 9:00:42 PM *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI Matt: I mean a neural model with increasingly complex features, as opposed to an algorithmic 3-D model (like video game graphics in reverse). Of course David rejects such ideas ( http://practicalai.org/Prize/Default.aspx ) even though the one proven working vision model uses it. Which is? and does what? (I'm starting to consider that vision and visual perception - or perhaps one should say common sense, since no sense in humans works independent of the others - may well be considerably *more* complex than language. The evolutionary time required to develop our common sense perception and conception of the world was vastly greater than that required to develop language. And we are as a culture merely in our babbling infancy in beginning to understand how sensory images work and are processed). *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
This is absolutely incredible. The answer was right there in the last paragraph: The present experiments suggest that the computation of object persistence appears to rely so heavily upon spatiotemporal information that it will not (or at least is unlikely to) use otherwise available surface feature information, particularly when there is conflicting spatiotemporal information. This reveals a striking limitation, given various theories that visual perception uses whatever shortcuts, or heuristics, it can to simplify processing, as well as the theory that perception evolves out of a buildup of the statistical nature of our environment (e.g., Purves Lotto, 2003). Instead, it appears that the object file system has “tunnel vision” and turns a blind eye to surface feature information, focusing on spatiotemporal information when computing persistence. So much for Matt's claim that the brain uses hierarchical features LOL Dave On Sat, Jul 24, 2010 at 11:52 PM, David Jones davidher...@gmail.com wrote: Check this out! The title Space and time, not surface features, guide object persistence says it all. http://pbr.psychonomic-journals.org/content/14/6/1199.full.pdf Over just the last couple days I have begun to realize that they are so right. My idea before of using high frame rates is also spot on. The brain does not use features as much as we think. First we construct a model of the object, then we probably decide what features to index it with for future search. If we know that the object occurs at a particular location in space, then we can learn a great deal about it with very little ambiguity! Of course, processing images at all is hard, but that's besides the point... The point is that we can automatically learn about the world using high frame rates and a simple heuristic for identifying specific objects in a scene. Because we can reliably identify them, we can learn an extremely large amount in a very short period of time. We can learn about how lighting affects the colors, noise, size, shape, components, attachment relationships, etc. etc. So, it is very likely that screenshots are not simpler than real images! lol. The objects in real images usually don't change as much, as drastically or as quickly as the objects in screenshots. That means that we can use the simple heuristics of size, shape, location and continuity of time to match objects and learn about them. Dave On Sat, Jul 24, 2010 at 9:10 PM, Matt Mahoney matmaho...@yahoo.comwrote: Mike Tintner wrote: Which is? The one right behind your eyes. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* Mike Tintner tint...@blueyonder.co.uk *To:* agi agi@v2.listbox.com *Sent:* Sat, July 24, 2010 9:00:42 PM *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI Matt: I mean a neural model with increasingly complex features, as opposed to an algorithmic 3-D model (like video game graphics in reverse). Of course David rejects such ideas ( http://practicalai.org/Prize/Default.aspx ) even though the one proven working vision model uses it. Which is? and does what? (I'm starting to consider that vision and visual perception - or perhaps one should say common sense, since no sense in humans works independent of the others - may well be considerably *more* complex than language. The evolutionary time required to develop our common sense perception and conception of the world was vastly greater than that required to develop language. And we are as a culture merely in our babbling infancy in beginning to understand how sensory images work and are processed). *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Clues to the Mind: Illusions / Vision
Yes. I think I may have discovered the keys to crack this puzzle wide open. The brain seems to use simplistic heuristics for depth perception and surface bounding. Once it has that, it can apply the spaciotemporal heuristic I mentioned in other emails to identify and track an object, which allows it to learn a lot with high confidence. So, that model would explain why we see depth perception illusions. Dave On Jul 25, 2010 1:04 AM, deepakjnath deepakjn...@gmail.com wrote: http://www.youtube.com/watch?v=QbKw0_v2clofeature=player_embedded What we see is not really what you see. Its what you see and what you know you are seeing. The brain superimposes the predicted images to the viewed image to actually have a perception of image. cheers, Deepak *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] Re: Huge Progress on the Core of AGI
An Update I think the following gets to the heart of general AI and what it takes to achieve it. It also provides us with evidence as to why general AI is so difficult. With this new knowledge in mind, I think I will be much more capable now of solving the problems and making it work. I've come to the conclusion lately that the best hypothesis is better because it is more predictive and then simpler than other hypotheses (in that order more predictive... then simpler). But, I am amazed at how difficult it is to quantitatively define more predictive and simpler for specific problems. This is why I have sometimes doubted the truth of the statement. In addition, the observations that the AI gets are not representative of all observations! This means that if your measure of predictiveness depends on the number of certain observations, it could make mistakes! So, the specific observations you are aware of may be unrepresentative of the predictiveness of a hypothesis relative to the truth. If you try to calculate which hypothesis is more predictive and you don't have the critical observations that would give you the right answer, you may get the wrong answer! This all depends of course on your method of calculation, which is quite elusive to define. Visual input from screenshots, for example, can be somewhat malicious. Things can move, appear, disappear or occlude each other suddenly. So, without sufficient knowledge it is hard to decide whether matches you find between such large changes are because it is the same object or a different object. This may indicate that bias and preprogrammed experience should be introduced to the AI before training. Either that or the training inputs should be carefully chosen to avoid malicious input and to make them nice for learning. This is the correspondence problem that is typical of computer vision and has never been properly solved. Such malicious input also makes it difficult to learn automatically because the AI doesn't have sufficient experience to know which changes or transformations are acceptable and which are not. It is immediately bombarded with malicious inputs. I've also realized that if a hypothesis is more explanatory, it may be better. But quantitatively defining explanatory is also elusive and truly depends on the specific problems you are applying it to because it is a heuristic. It is not a true measure of correctness. It is not loyal to the truth. More explanatory is really a heuristic that helps us find hypothesis that are more predictive. The true measure of whether a hypothesis is better is simply the most accurate and predictive hypothesis. That is the ultimate and true measure of correctness. Also, since we can't measure every possible prediction or every last prediction (and we certainly can't predict everything), our measure of predictiveness can't possibly be right all the time! We have no choice but to use a heuristic of some kind. So, its clear to me that the right hypothesis is more predictive and then simpler. But, it is also clear that there will never be a single measure of this that can be applied to all problems. I hope to eventually find a nice model for how to apply it to different problems though. This may be the reason that so many people have tried and failed to develop general AI. Yes, there is a solution. But there is no silver bullet that can be applied to all problems. Some methods are better than others. But I think another major reason of the failures is that people think they can predict things without sufficient information. By approaching the problem this way, we compound the need for heuristics and the errors they produce because we simply don't have sufficient information to make a good decision with limited evidence. If approached correctly, the right solution would solve many more problems with the same efforts than a poor solution would. It would also eliminate some of the difficulties we currently face if sufficient data is available to learn from. In addition to all this theory about better hypotheses, you have to add on the need to solve problems in reasonable time. This also compounds the difficulty of the problem and the complexity of solutions. I am always fascinated by the extraordinary difficulty and complexity of this problem. The more I learn about it, the more I appreciate it. Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
Because simpler is not better if it is less predictive. On Thu, Jul 22, 2010 at 1:21 PM, Abram Demski abramdem...@gmail.com wrote: Jim, Why more predictive *and then* simpler? --Abram On Thu, Jul 22, 2010 at 11:49 AM, David Jones davidher...@gmail.comwrote: An Update I think the following gets to the heart of general AI and what it takes to achieve it. It also provides us with evidence as to why general AI is so difficult. With this new knowledge in mind, I think I will be much more capable now of solving the problems and making it work. I've come to the conclusion lately that the best hypothesis is better because it is more predictive and then simpler than other hypotheses (in that order more predictive... then simpler). But, I am amazed at how difficult it is to quantitatively define more predictive and simpler for specific problems. This is why I have sometimes doubted the truth of the statement. In addition, the observations that the AI gets are not representative of all observations! This means that if your measure of predictiveness depends on the number of certain observations, it could make mistakes! So, the specific observations you are aware of may be unrepresentative of the predictiveness of a hypothesis relative to the truth. If you try to calculate which hypothesis is more predictive and you don't have the critical observations that would give you the right answer, you may get the wrong answer! This all depends of course on your method of calculation, which is quite elusive to define. Visual input from screenshots, for example, can be somewhat malicious. Things can move, appear, disappear or occlude each other suddenly. So, without sufficient knowledge it is hard to decide whether matches you find between such large changes are because it is the same object or a different object. This may indicate that bias and preprogrammed experience should be introduced to the AI before training. Either that or the training inputs should be carefully chosen to avoid malicious input and to make them nice for learning. This is the correspondence problem that is typical of computer vision and has never been properly solved. Such malicious input also makes it difficult to learn automatically because the AI doesn't have sufficient experience to know which changes or transformations are acceptable and which are not. It is immediately bombarded with malicious inputs. I've also realized that if a hypothesis is more explanatory, it may be better. But quantitatively defining explanatory is also elusive and truly depends on the specific problems you are applying it to because it is a heuristic. It is not a true measure of correctness. It is not loyal to the truth. More explanatory is really a heuristic that helps us find hypothesis that are more predictive. The true measure of whether a hypothesis is better is simply the most accurate and predictive hypothesis. That is the ultimate and true measure of correctness. Also, since we can't measure every possible prediction or every last prediction (and we certainly can't predict everything), our measure of predictiveness can't possibly be right all the time! We have no choice but to use a heuristic of some kind. So, its clear to me that the right hypothesis is more predictive and then simpler. But, it is also clear that there will never be a single measure of this that can be applied to all problems. I hope to eventually find a nice model for how to apply it to different problems though. This may be the reason that so many people have tried and failed to develop general AI. Yes, there is a solution. But there is no silver bullet that can be applied to all problems. Some methods are better than others. But I think another major reason of the failures is that people think they can predict things without sufficient information. By approaching the problem this way, we compound the need for heuristics and the errors they produce because we simply don't have sufficient information to make a good decision with limited evidence. If approached correctly, the right solution would solve many more problems with the same efforts than a poor solution would. It would also eliminate some of the difficulties we currently face if sufficient data is available to learn from. In addition to all this theory about better hypotheses, you have to add on the need to solve problems in reasonable time. This also compounds the difficulty of the problem and the complexity of solutions. I am always fascinated by the extraordinary difficulty and complexity of this problem. The more I learn about it, the more I appreciate it. Dave *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- Abram Demski http://lo-tho.blogspot.com/ http://groups.google.com/group
Re: [agi] Re: Huge Progress on the Core of AGI
It's certainly not as simple as you claim. First, assigning a probability is not always possible, nor is it easy. The factors in calculating that probability are unknown and are not the same for every instance. Since we do not know what combination of observations we will see, we cannot have a predefined set of probabilities, nor is it any easier to create a probability function that generates them for us. That is just as exactly what I meant by quantitatively define the predictiveness... it would be proportional to the probability. Second, if you can define a program ina way that is always simpler when it is smaller, then you can do the same thing without a program. I don't think it makes any sense to do it this way. It is not that simple. If it was, we could solve a large portion of agi easily. On Thu, Jul 22, 2010 at 3:16 PM, Matt Mahoney matmaho...@yahoo.com wrote: David Jones wrote: But, I am amazed at how difficult it is to quantitatively define more predictive and simpler for specific problems. It isn't hard. To measure predictiveness, you assign a probability to each possible outcome. If the actual outcome has probability p, you score a penalty of log(1/p) bits. To measure simplicity, use the compressed size of the code for your prediction algorithm. Then add the two scores together. That's how it is done in the Calgary challenge http://www.mailcom.com/challenge/ and in my own text compression benchmark. -- Matt Mahoney, matmaho...@yahoo.com *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Thu, July 22, 2010 3:11:46 PM *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI Because simpler is not better if it is less predictive. On Thu, Jul 22, 2010 at 1:21 PM, Abram Demski abramdem...@gmail.com wrote: Jim, Why more predictive *and then* simpler? --Abram On Thu, Jul 22, 2010 at 11:49 AM, David Jones davidher...@gmail.com wrote: An Update I think the following gets to the heart of general AI and what it takes to achieve it. It also provides us with evidence as to why general AI is so difficult. With this new knowledge in mind, I think I will be much more capable now of solving the problems and making it work. I've come to the conclusion lately that the best hypothesis is better because it is more predictive and then simpler than other hypotheses (in that order more predictive... then simpler). But, I am amazed at how difficult it is to quantitatively define more predictive and simpler for specific problems. This is why I have sometimes doubted the truth of the statement. In addition, the observations that the AI gets are not representative of all observations! This means that if your measure of predictiveness depends on the number of certain observations, it could make mistakes! So, the specific observations you are aware of may be unrepresentative of the predictiveness of a hypothesis relative to the truth. If you try to calculate which hypothesis is more predictive and you don't have the critical observations that would give you the right answer, you may get the wrong answer! This all depends of course on your method of calculation, which is quite elusive to define. Visual input from screenshots, for example, can be somewhat malicious. Things can move, appear, disappear or occlude each other suddenly. So, without sufficient knowledge it is hard to decide whether matches you find between such large changes are because it is the same object or a different object. This may indicate that bias and preprogrammed experience should be introduced to the AI before training. Either that or the training inputs should be carefully chosen to avoid malicious input and to make them nice for learning. This is the correspondence problem that is typical of computer vision and has never been properly solved. Such malicious input also makes it difficult to learn automatically because the AI doesn't have sufficient experience to know which changes or transformations are acceptable and which are not. It is immediately bombarded with malicious inputs. I've also realized that if a hypothesis is more explanatory, it may be better. But quantitatively defining explanatory is also elusive and truly depends on the specific problems you are applying it to because it is a heuristic. It is not a true measure of correctness. It is not loyal to the truth. More explanatory is really a heuristic that helps us find hypothesis that are more predictive. The true measure of whether a hypothesis is better is simply the most accurate and predictive hypothesis. That is the ultimate and true measure of correctness. Also, since we can't measure every possible prediction or every last prediction (and we certainly can't predict everything), our measure of predictiveness can't possibly be right all the time! We have no choice but to use a heuristic of some kind. So, its clear to me that the right hypothesis is more predictive and then simpler. But, it is also clear
Re: [agi] Of definitions and tests of AGI
Training data is not available in many real problems. I don't think training data should be used as the main learning mechanism. It likely won't solve any of the problems. On Jul 21, 2010 2:52 AM, deepakjnath deepakjn...@gmail.com wrote: Yes we could do a 4x4 tic tac toe game like this in a PC. The training sets can be generated simply by playing the agents against each other using random moves and letting the agents know if it passed or failed as a feedback mechanism. Cheers, Deepak On Wed, Jul 21, 2010 at 9:02 AM, Matt Mahoney matmaho...@yahoo.com wrote: Mike, I think we a... -- cheers, Deepak *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Is there any Contest or test to ensure that a System is AGI?
not really. On Sun, Jul 18, 2010 at 9:41 AM, deepakjnath deepakjn...@gmail.com wrote: Yes, but is there a competition like the XPrize or something that we can work towards. ? On Sun, Jul 18, 2010 at 6:40 PM, Panu Horsmalahti nawi...@gmail.comwrote: 2010/7/18 deepakjnath deepakjn...@gmail.com I wanted to know if there is any bench mark test that can really convince majority of today's AGIers that a System is true AGI? Is there some real prize like the XPrize for AGI or AI in general? thanks, Deepak Have you heard about the Turing test? - Panu Horsmalahti *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- cheers, Deepak *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Of definitions and tests of AGI
If you can't convince someone, clearly something is wrong with it. I don't think a test is the right way to do this. Which is why I haven't commented much. When you understand how to create AGI, it will be obvious that it is AGI or that it is what you intend it to be. You'll then understand how what you have built fits into the bigger scheme of things. There is no such point at which you can say something is AGI and not AGI. Intelligence is a very subjective thing that really depends on your goals. Someone will always say it is not good enough. But if it really works, people will quickly realize it based on results. What you want is to develop a system that can learn about the world or its environment in a general way so that it can solve arbitrary problems, be able to plan in general ways, act in general ways and perform the types of goals you want it to perform. Dave On Sun, Jul 18, 2010 at 3:03 PM, deepakjnath deepakjn...@gmail.com wrote: So if I have a system that is close to AGI, I have no way of really knowing it right? Even if I believe that my system is a true AGI there is no way of convincing the others irrefutably that this system is indeed a AGI not just an advanced AI system. I have read the toy box problem and rock wall problem, but not many people will still be convinced I am sure. I wanted to know that if there is any consensus on a general problem which can be solved and only solved by a true AGI. Without such a test bench how will we know if we are moving closer or away from our quest. There is no map. Deepak On Sun, Jul 18, 2010 at 11:50 PM, Mike Tintner tint...@blueyonder.co.ukwrote: I realised that what is needed is a *joint* definition *and* range of tests of AGI. Benamin Johnston has submitted one valid test - the toy box problem. (See archives). I have submitted another still simpler valid test - build a rock wall from rocks given, (or fill an earth hole with rocks). However, I see that there are no valid definitions of AGI that explain what AGI is generally , and why these tests are indeed AGI. Google - there are v. few defs. of AGI or Strong AI, period. The most common: AGI is human-level intelligence - is an embarrassing non-starter - what distinguishes human intelligence? No explanation offered. The other two are also inadequate if not as bad: Ben's solves a variety of complex problems in a variety of complex environments. Nope, so does a multitasking narrow AI. Complexity does not distinguish AGI. Ditto Pei's - something to do with insufficient knowledge and resources... Insufficient is open to narrow AI interpretations and reducible to mathematically calculable probabilities.or uncertainties. That doesn't distinguish AGI from narrow AI. The one thing we should all be able to agree on (but who can be sure?) is that: ** an AGI is a general intelligence system, capable of independent learning** i.e. capable of independently learning new activities/skills with minimal guidance or even, ideally, with zero guidance (as humans and animals are) - and thus acquiring a general, all-round range of intelligence.. This is an essential AGI goal - the capacity to keep entering and mastering new domains of both mental and physical skills WITHOUT being specially programmed each time - that crucially distinguishes it from narrow AI's, which have to be individually programmed anew for each new task. Ben's AGI dog exemplified this in a v simple way - the dog is supposed to be able to learn to fetch a ball, with only minimal instructions, as real dogs do - they can learn a whole variety of new skills with minimal instruction. But I am confident Ben's dog can't actually do this. However, the independent learning def. while focussing on the distinctive AGI goal, still is not detailed enough by itself. It requires further identification of the **cognitive operations** which distinguish AGI, and wh. are exemplified by the above tests. [I'll stop there for interruptions/comments continue another time]. P.S. Deepakjnath, It is vital to realise that the overwhelming majority of AGI-ers do not * want* an AGI test - Ben has never gone near one, and is merely typical in this respect. I'd put almost all AGI-ers here in the same league as the US banks, who only want mark-to-fantasy rather than mark-to-market tests of their assets. *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- cheers, Deepak *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed:
Re: [agi] Of definitions and tests of AGI
Deepak, I think you would be much better off focusing on something more practical. Understanding a movie and all the myriad things going on, their significance, etc... that's AI complete. There is no way you are going to get there without a hell of a lot of steps in between. So, you might as well focus on the steps required to get there. Such a test is so complicated, that you cannot even start, except to look for simpler test cases and goals. My approach to testing agi has been to define what AGI must accomplish. Which I have in the following steps: 1) understand the environment 2) understand ones own actions and how they affect the environment 3) understand language 4) learn goals from other people through language 5) perform planning and attempt to achieve goals 6) other miscellaneous requirements. Each step must be accomplished in a general way. By general, I mean that it can solve many many problems with the same programming. Each step must be done in order because each step requires previous steps to proceed. So, to me, the most important place to start is general environment understanding. Then, now that you know where to start, you pick more specific goals and test cases. How do you develop and test general environment understanding? What is a simple test case you can develop on? What are the fundamental problems and principles involved? What is required to solve these problems? Those are the sorts of tests you should be considering. But that only comes after you decide what AGI requires and steps required. Maybe you'll agree with me, maybe you won't. So, that's how I would recommend going about it. Dave On Sun, Jul 18, 2010 at 4:04 PM, deepakjnath deepakjn...@gmail.com wrote: Let me clarify. As you all know there are somethings computers are good at doing and somethings that Humans can do but a computer cannot. One of the test that I was thinking about recently is to have to movies show to the AGI. Both movies will have the same story but it would be a totally different remake of the film probably in different languages and settings. If the AGI is able to understand the sub plot and say that the story line is similar in the two movies then it could be a good test for AGI structure. The ability of a system to understand its environment and underlying sub plots is an important requirement of AGI. Deepak On Mon, Jul 19, 2010 at 1:14 AM, Mike Tintner tint...@blueyonder.co.ukwrote: Please explain/expound freely why you're not convinced - and indicate what you expect, - and I'll reply - but it may not be till tomorrow. Re your last point, there def. is no consensus on a general problem/test OR a def. of AGI. One flaw in your expectations seems to be a desire for a single test - almost by definition, there is no such thing as a) a single test - i.e. there should be at least a dual or serial test - having passed any given test, like the rock/toy test, the AGI must be presented with a new adjacent test for wh. it has had no preparation, like say building with cushions or sand bags or packing with fruit. (and neither rock/toy test state that clearly) b) one kind of test - this is an AGI, so it should be clear that if it can pass one kind of test, it has the basic potential to go on to many different kinds, and it doesn't really matter which kind of test you start with - that is partly the function of having a good.definition of AGI . *From:* deepakjnath deepakjn...@gmail.com *Sent:* Sunday, July 18, 2010 8:03 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] Of definitions and tests of AGI So if I have a system that is close to AGI, I have no way of really knowing it right? Even if I believe that my system is a true AGI there is no way of convincing the others irrefutably that this system is indeed a AGI not just an advanced AI system. I have read the toy box problem and rock wall problem, but not many people will still be convinced I am sure. I wanted to know that if there is any consensus on a general problem which can be solved and only solved by a true AGI. Without such a test bench how will we know if we are moving closer or away from our quest. There is no map. Deepak On Sun, Jul 18, 2010 at 11:50 PM, Mike Tintner tint...@blueyonder.co.ukwrote: I realised that what is needed is a *joint* definition *and* range of tests of AGI. Benamin Johnston has submitted one valid test - the toy box problem. (See archives). I have submitted another still simpler valid test - build a rock wall from rocks given, (or fill an earth hole with rocks). However, I see that there are no valid definitions of AGI that explain what AGI is generally , and why these tests are indeed AGI. Google - there are v. few defs. of AGI or Strong AI, period. The most common: AGI is human-level intelligence - is an embarrassing non-starter - what distinguishes human intelligence? No explanation offered. The other two are also inadequate
Re: [agi] Is there any Contest or test to ensure that a System is AGI?
Ian, Although most people see natural language as one of the most important parts of AGI, if you think about it carefully, you'll realize that solving natural language could be done with sufficient knowledge of the world and sufficient ability to learn this knowledge automatically. That's why i don't consider natural language a problem we can focus on until we solve the knowledge problem... which is what I'm focusing on. Dave 2010/7/18 Ian Parker ianpark...@gmail.com In my view the main obstacle to AGI is the understanding of Natural Language. If we have NL comprehension we have the basis for doing a whole host of marvellous things. There is the Turing test. A good question to ask is What is the difference between laying concrete at 50C and fighting Israel. Google translated wsT jw AlmErkp or وسط جو المعركة as central air battle. Correct is the climatic environmental battle or a more free translation would be the battle against climate and environment. In Turing competitions no one ever asks the questions that really would tell AGI apart from a brand X chatterbox. http://sites.google.com/site/aitranslationproject/Home/formalmethods http://sites.google.com/site/aitranslationproject/Home/formalmethodsWe can I think say that anything which can carry out the program of my blog would be well on its way. AGI will also be the link between NL and formal mathematics. Let me take yet another example. http://sites.google.com/site/aitranslationproject/deepknowled Google translated it as 4 times the temperature. Ponder this, you have in fact 3 chances to get this right. 1) درجة means degree. GT has not translated this word. In this context it means power. 2) If you search for Stefan Boltzmann or Black Body Google gives you the correct law. 3) The translation is obviously mathematically incorrect from the dimensional stand-point. This 3 things in fact represent different aspects of knowledge. In AGI they all have to be present. The other interesting point is that there are programs in existence now that will address the last two questions. A translator that produces OWL solves 2. If we match up AGI to Mizar we can put dimensions into the proof engine. There are a great many things on the Web which will solve specific problems. NL is *THE* problem since it will allow navigation between the different programs on the Web. MOLTO BTW does have its mathematical parts even though it is primerally billed as a translator. - Ian Parker On 18 July 2010 14:41, deepakjnath deepakjn...@gmail.com wrote: Yes, but is there a competition like the XPrize or something that we can work towards. ? On Sun, Jul 18, 2010 at 6:40 PM, Panu Horsmalahti nawi...@gmail.comwrote: 2010/7/18 deepakjnath deepakjn...@gmail.com I wanted to know if there is any bench mark test that can really convince majority of today's AGIers that a System is true AGI? Is there some real prize like the XPrize for AGI or AI in general? thanks, Deepak Have you heard about the Turing test? - Panu Horsmalahti *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- cheers, Deepak *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Is there any Contest or test to ensure that a System is AGI?
Oh, I wanted to add one thing that I've learned recently. The core problem of AGI is to come up with hypotheses (hopefully the right hypothesis or one that is good enough is included) and then determine whether the hypothesis is 1) acceptable and 2) better than other hypotheses. In addition, you have to have a way to decide *when* to look for better hypotheses, because you can't just always be looking at all possible hypotheses. So, with that in mind, the reason that natural language can only be very roughly approximated without a lot more knowledge is because there isn't sufficient knowledge to say that one hypothesis is better than another in the vast majority of cases. The AI doesn't have sufficient *reason* to think that the right hypothesis is better than others. The only way to give it that sufficient reason is to give it sufficient knowledge. Dave 2010/7/18 David Jones davidher...@gmail.com Ian, Although most people see natural language as one of the most important parts of AGI, if you think about it carefully, you'll realize that solving natural language could be done with sufficient knowledge of the world and sufficient ability to learn this knowledge automatically. That's why i don't consider natural language a problem we can focus on until we solve the knowledge problem... which is what I'm focusing on. Dave 2010/7/18 Ian Parker ianpark...@gmail.com In my view the main obstacle to AGI is the understanding of Natural Language. If we have NL comprehension we have the basis for doing a whole host of marvellous things. There is the Turing test. A good question to ask is What is the difference between laying concrete at 50C and fighting Israel. Google translated wsT jw AlmErkp or وسط جو المعركة as central air battle. Correct is the climatic environmental battle or a more free translation would be the battle against climate and environment. In Turing competitions no one ever asks the questions that really would tell AGI apart from a brand X chatterbox. http://sites.google.com/site/aitranslationproject/Home/formalmethods http://sites.google.com/site/aitranslationproject/Home/formalmethodsWe can I think say that anything which can carry out the program of my blog would be well on its way. AGI will also be the link between NL and formal mathematics. Let me take yet another example. http://sites.google.com/site/aitranslationproject/deepknowled Google translated it as 4 times the temperature. Ponder this, you have in fact 3 chances to get this right. 1) درجة means degree. GT has not translated this word. In this context it means power. 2) If you search for Stefan Boltzmann or Black Body Google gives you the correct law. 3) The translation is obviously mathematically incorrect from the dimensional stand-point. This 3 things in fact represent different aspects of knowledge. In AGI they all have to be present. The other interesting point is that there are programs in existence now that will address the last two questions. A translator that produces OWL solves 2. If we match up AGI to Mizar we can put dimensions into the proof engine. There are a great many things on the Web which will solve specific problems. NL is *THE* problem since it will allow navigation between the different programs on the Web. MOLTO BTW does have its mathematical parts even though it is primerally billed as a translator. - Ian Parker On 18 July 2010 14:41, deepakjnath deepakjn...@gmail.com wrote: Yes, but is there a competition like the XPrize or something that we can work towards. ? On Sun, Jul 18, 2010 at 6:40 PM, Panu Horsmalahti nawi...@gmail.comwrote: 2010/7/18 deepakjnath deepakjn...@gmail.com I wanted to know if there is any bench mark test that can really convince majority of today's AGIers that a System is true AGI? Is there some real prize like the XPrize for AGI or AI in general? thanks, Deepak Have you heard about the Turing test? - Panu Horsmalahti *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- cheers, Deepak *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] NL parsing
This is actually a great example of why we should not try to write AGI as something able to solve any possible problem generally. We, strong ai agents, are not able to understand this sentence without quite a lot more information. Likewise, we shouldn't expect a general AI to try many possibilities until it is able to solve such a maliciously constructed sentence. There isn't explanatory reason to believe most of the possible hypotheses. We need more information to come up with possible hypotheses, which we can then test out on the sentence and confirm. That' why our additional knowledge from the blog is the only way we can reasonably disambiguate the sentence. Normal natural language disambiguation is similar in that way. Dave On Fri, Jul 16, 2010 at 11:29 AM, Matt Mahoney matmaho...@yahoo.com wrote: That that that Buffalo buffalo that Buffalo buffalo buffalo buffalo that Buffalo buffalo that Buffalo buffalo buffalo. -- Matt Mahoney, matmaho...@yahoo.com - Original Message From: Mike Tintner tint...@blueyonder.co.uk To: agi agi@v2.listbox.com Sent: Fri, July 16, 2010 11:05:51 AM Subject: Re: [agi] NL parsing Or if you want to be pedantic about caps, the speaker is identifying 3 buffaloes from Buffalo, 2 from elsewhere. Anyone got any other readings? -- From: Jiri Jelinek jjelinek...@gmail.com Sent: Friday, July 16, 2010 3:12 PM To: agi agi@v2.listbox.com Subject: [agi] NL parsing Believe it or not, this sentence is grammatically correct and has meaning: 'Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo.' source: http://www.mentalfloss.com/blogs/archives/13120 :-) --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] NL parsing
Mike, Your reading requires extensive knowledge also to even think that it is explanatory. You've heard people say a noun over and over to note that they saw one. So, that is the only reason you are able to try to disambiguate the sentence this way. Even so, without context, the sentence is still not very explanatory, and so, any regular person would look for more information because it seems to be such a strange sentence. Dave On Fri, Jul 16, 2010 at 12:04 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Dave: That's why our additional knowledge from the blog is the only way we can reasonably disambiguate the sentence. Contradicted by my reading. The particular blog reading was esoteric sure. But you do have to be capable of creative readings as humans are - that's the fundamental challenge of language. But of course no machine understands language yet, period - and isn't likely to for a v. v. long time. *From:* David Jones davidher...@gmail.com *Sent:* Friday, July 16, 2010 4:35 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] NL parsing This is actually a great example of why we should not try to write AGI as something able to solve any possible problem generally. We, strong ai agents, are not able to understand this sentence without quite a lot more information. Likewise, we shouldn't expect a general AI to try many possibilities until it is able to solve such a maliciously constructed sentence. There isn't explanatory reason to believe most of the possible hypotheses. We need more information to come up with possible hypotheses, which we can then test out on the sentence and confirm. That' why our additional knowledge from the blog is the only way we can reasonably disambiguate the sentence. Normal natural language disambiguation is similar in that way. Dave On Fri, Jul 16, 2010 at 11:29 AM, Matt Mahoney matmaho...@yahoo.comwrote: That that that Buffalo buffalo that Buffalo buffalo buffalo buffalo that Buffalo buffalo that Buffalo buffalo buffalo. -- Matt Mahoney, matmaho...@yahoo.com - Original Message From: Mike Tintner tint...@blueyonder.co.uk To: agi agi@v2.listbox.com Sent: Fri, July 16, 2010 11:05:51 AM Subject: Re: [agi] NL parsing Or if you want to be pedantic about caps, the speaker is identifying 3 buffaloes from Buffalo, 2 from elsewhere. Anyone got any other readings? -- From: Jiri Jelinek jjelinek...@gmail.com Sent: Friday, July 16, 2010 3:12 PM To: agi agi@v2.listbox.com Subject: [agi] NL parsing Believe it or not, this sentence is grammatically correct and has meaning: 'Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo.' source: http://www.mentalfloss.com/blogs/archives/13120 :-) --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How do we Score Hypotheses?
It is no wonder that I'm having a hard time finding documentation on hypothesis scoring. Few can agree on how to do it and there is much debate about it. I noticed though that a big reason for the problems is that explanatory reasoning is being applied to many diverse problems. I think, like I mentioned before, that people should not try to come up with a single universal rule set for applying explanatory reasoning to every possible problem. So, maybe that's where the hold up is. I've been testing my ideas out on complex examples. But now I'm going to go back to simplified model testing (although not as simple as black squares :) ) and work my way up again. Dave On Wed, Jul 14, 2010 at 12:59 PM, David Jones davidher...@gmail.com wrote: Actually, I just realized that there is a way to included inductive knowledge and experience into this algorithm. Inductive knowledge and experience about a specific object or object type can be exploited to know which hypotheses in the past were successful, and therefore which hypothesis is most likely. By choosing the most likely hypothesis first, we skip a lot of messy hypothesis comparison processing and analysis. If we choose the right hypothesis first, all we really have to do is verify that this hypothesis reveals in the data what we expect to be there. If we confirm what we expect, that is reason enough not to look for other hypotheses because the data is explained by what we originally believed to be likely. We only look for additional hypotheses when we find something unexplained. And even then, we don't look at the whole problem. We only look at what we have to to explain the unexplained data. In fact, we could even ignore the unexplained data if we believe, from experience, that it isn't pertinent. I discovered this because I'm analyzing how a series of hypotheses are navigated when analyzing images. It seems to me that it is done very similarly to way we do it. We sort of confirm what we expect and try to explain what we don't expect. We try out hypotheses in a sort of trial and error manor and see how each hypothesis affects what we find in the image. If we confirm things because of the hypothesis, we are likely to keep it. We keep going, navigating the tree of hypotheses, conflicts and unexpected observations until we find a good hypothesis. Something like that. I'm attempting to construct an algorithm for doing this as I analyze specific problems. Dave On Wed, Jul 14, 2010 at 10:22 AM, David Jones davidher...@gmail.comwrote: What do you mean by definitive events? I guess the first problem I see with my approach is that the movement of the window is also a hypothesis. I need to analyze it in more detail and see how the tree of hypotheses affects the hypotheses regarding the es on the windows. What I believe is that these problems can be broken down into types of hypotheses, types of events and types of relationships. then those types can be reasoned about in a general way. If possible, then you have a method for reasoning about any object that is covered by the types of hypotheses, events and relationships that you have defined. How to reason about specific objects should not be preprogrammed. But, I think the solution to this part of AGI is to find general ways to reason about a small set of concepts that can be combined to describe specific objects and situations. There are other parts to AGI that I am not considering yet. I believe the problem has to be broken down into separate pieces and understood before putting it back together into a complete system. I have not covered inductive learning for example, which would be an important part of AGI. I have also not yet incorporated learned experience into the algorithm, which is also important. The general AI problem is way too complicated to consider all at once. I simply can't solve hypothesis generation, comparison and disambiguation while at the same time solving induction and experience-based reasoning. It becomes unwieldly. So, I'm starting where I can and I'll work my way up to the full complexity of the problem. I don't really understand what you mean here: The central unsolved problem, in my view, is: How can hypotheses be conceptually integrated along with the observable definitive events of the problem to form good explanatory connections that can mesh well with other knowledge about the problem that is considered to be reliable. The second problem is finding efficient ways to represent this complexity of knowledge so that the program can utilize it efficiently. You also might want to include concrete problems to analyze for your central problem suggestions. That would help define the problem a bit better for analysis. Dave On Wed, Jul 14, 2010 at 8:30 AM, Jim Bromer jimbro...@gmail.com wrote: On Tue, Jul 13, 2010 at 9:05 PM, Jim Bromer jimbro...@gmail.com wrote: Even if you refined your model until it was just
Re: [agi] How do we Score Hypotheses?
:) You say that as if bayesian explanatory reasoning is the only way. There is much debate over bayesian explanatory reasoning and non-bayesian. There are pros and cons to bayesian methods. Likewise, there is the problem with non-bayesian methods because few have figured out how to do it effectively. I'm still going to pursue a non-bayesian approach because I believe there is likely more merit to it and that the short-comings can be overcome. Dave On Thu, Jul 15, 2010 at 10:54 AM, Matt Mahoney matmaho...@yahoo.com wrote: Hypotheses are scored using Bayes law. Let D be your observed data and H be your hypothesis. Then p(H|D) = p(D|H)p(H)/p(D). Since p(D) is constant, you can remove it and rank hypotheses by p(D|H)p(H). p(H) can be estimated using the minimum description length principle or Solomonoff induction. Ideally, p(H) = 2^-|H| where |H| is the length (in bits) of the description of the hypothesis. The value is language dependent, so this method is not perfect. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Thu, July 15, 2010 10:22:44 AM *Subject:* Re: [agi] How do we Score Hypotheses? It is no wonder that I'm having a hard time finding documentation on hypothesis scoring. Few can agree on how to do it and there is much debate about it. I noticed though that a big reason for the problems is that explanatory reasoning is being applied to many diverse problems. I think, like I mentioned before, that people should not try to come up with a single universal rule set for applying explanatory reasoning to every possible problem. So, maybe that's where the hold up is. I've been testing my ideas out on complex examples. But now I'm going to go back to simplified model testing (although not as simple as black squares :) ) and work my way up again. Dave On Wed, Jul 14, 2010 at 12:59 PM, David Jones davidher...@gmail.comwrote: Actually, I just realized that there is a way to included inductive knowledge and experience into this algorithm. Inductive knowledge and experience about a specific object or object type can be exploited to know which hypotheses in the past were successful, and therefore which hypothesis is most likely. By choosing the most likely hypothesis first, we skip a lot of messy hypothesis comparison processing and analysis. If we choose the right hypothesis first, all we really have to do is verify that this hypothesis reveals in the data what we expect to be there. If we confirm what we expect, that is reason enough not to look for other hypotheses because the data is explained by what we originally believed to be likely. We only look for additional hypotheses when we find something unexplained. And even then, we don't look at the whole problem. We only look at what we have to to explain the unexplained data. In fact, we could even ignore the unexplained data if we believe, from experience, that it isn't pertinent. I discovered this because I'm analyzing how a series of hypotheses are navigated when analyzing images. It seems to me that it is done very similarly to way we do it. We sort of confirm what we expect and try to explain what we don't expect. We try out hypotheses in a sort of trial and error manor and see how each hypothesis affects what we find in the image. If we confirm things because of the hypothesis, we are likely to keep it. We keep going, navigating the tree of hypotheses, conflicts and unexpected observations until we find a good hypothesis. Something like that. I'm attempting to construct an algorithm for doing this as I analyze specific problems. Dave On Wed, Jul 14, 2010 at 10:22 AM, David Jones davidher...@gmail.comwrote: What do you mean by definitive events? I guess the first problem I see with my approach is that the movement of the window is also a hypothesis. I need to analyze it in more detail and see how the tree of hypotheses affects the hypotheses regarding the es on the windows. What I believe is that these problems can be broken down into types of hypotheses, types of events and types of relationships. then those types can be reasoned about in a general way. If possible, then you have a method for reasoning about any object that is covered by the types of hypotheses, events and relationships that you have defined. How to reason about specific objects should not be preprogrammed. But, I think the solution to this part of AGI is to find general ways to reason about a small set of concepts that can be combined to describe specific objects and situations. There are other parts to AGI that I am not considering yet. I believe the problem has to be broken down into separate pieces and understood before putting it back together into a complete system. I have not covered inductive learning for example, which would be an important part of AGI. I have also not yet incorporated
Re: [agi] How do we Score Hypotheses?
Jim, even that isn't an obvious event. You don't know what is background and what is not. You don't even know if there is an object or not. You don't know if anything moved or not. You can make some observations using predefined methods and then see if you find matches... then hypothesize about the matches... It all has to be learned and figured out through reasoning. That's why I asked what you meant by definitive events. Nothing is really definitive. It is all hypothesized in a non-monotonic manner. Dave On Thu, Jul 15, 2010 at 12:01 PM, Jim Bromer jimbro...@gmail.com wrote: On Wed, Jul 14, 2010 at 10:22 AM, David Jones davidher...@gmail.comwrote: What do you mean by definitive events? I was just trying to find a way to designate obsverations that would be reliably obvious to a computer program. This has something to do with the assumptions that you are using. For example if some object appeared against a stable background and it was a different color than the background, it would be a definitive observation event because your algorithm could detect it with some certainty and use it in the definition of other more complicated events (like occlusion.) Notice that this example would not necessarily be so obvious (a definitive event) using a camera, because there are a number of ways that an illusion (of some kind) could end up as a data event. I will try to reply to the rest of your message sometime later. Jim Bromer *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How do we Score Hypotheses?
On screenshots, the point of view is equivalent to the absolute positions and their relative positions using absolute(screen x and y) measurements. You don't need a robot to learn about how AGI works and figure out how to solve some problems. It would be a terrible mistake to spend years, or even weeks for that matter, on robotics before getting started. Dave On Thu, Jul 15, 2010 at 1:09 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Sounds like a good explanation of why a body is essential for vision - not just for POV and orientation [up/left/right/down/ towards/ away] but for comparison and yardstick - you do know when your body or parts thereof are moving -and it's not merely touch but the comparison of other objects still and moving with your own moving hands and body that is important. The more you go into it, the crazier the prospect of vision without eyes in a body becomes. *From:* David Jones davidher...@gmail.com *Sent:* Thursday, July 15, 2010 5:54 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] How do we Score Hypotheses? Jim, even that isn't an obvious event. You don't know what is background and what is not. You don't even know if there is an object or not. You don't know if anything moved or not. You can make some observations using predefined methods and then see if you find matches... then hypothesize about the matches... It all has to be learned and figured out through reasoning. That's why I asked what you meant by definitive events. Nothing is really definitive. It is all hypothesized in a non-monotonic manner. Dave On Thu, Jul 15, 2010 at 12:01 PM, Jim Bromer jimbro...@gmail.com wrote: On Wed, Jul 14, 2010 at 10:22 AM, David Jones davidher...@gmail.comwrote: What do you mean by definitive events? I was just trying to find a way to designate obsverations that would be reliably obvious to a computer program. This has something to do with the assumptions that you are using. For example if some object appeared against a stable background and it was a different color than the background, it would be a definitive observation event because your algorithm could detect it with some certainty and use it in the definition of other more complicated events (like occlusion.) Notice that this example would not necessarily be so obvious (a definitive event) using a camera, because there are a number of ways that an illusion (of some kind) could end up as a data event. I will try to reply to the rest of your message sometime later. Jim Bromer *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How do we Score Hypotheses?
What do you mean by definitive events? I guess the first problem I see with my approach is that the movement of the window is also a hypothesis. I need to analyze it in more detail and see how the tree of hypotheses affects the hypotheses regarding the es on the windows. What I believe is that these problems can be broken down into types of hypotheses, types of events and types of relationships. then those types can be reasoned about in a general way. If possible, then you have a method for reasoning about any object that is covered by the types of hypotheses, events and relationships that you have defined. How to reason about specific objects should not be preprogrammed. But, I think the solution to this part of AGI is to find general ways to reason about a small set of concepts that can be combined to describe specific objects and situations. There are other parts to AGI that I am not considering yet. I believe the problem has to be broken down into separate pieces and understood before putting it back together into a complete system. I have not covered inductive learning for example, which would be an important part of AGI. I have also not yet incorporated learned experience into the algorithm, which is also important. The general AI problem is way too complicated to consider all at once. I simply can't solve hypothesis generation, comparison and disambiguation while at the same time solving induction and experience-based reasoning. It becomes unwieldly. So, I'm starting where I can and I'll work my way up to the full complexity of the problem. I don't really understand what you mean here: The central unsolved problem, in my view, is: How can hypotheses be conceptually integrated along with the observable definitive events of the problem to form good explanatory connections that can mesh well with other knowledge about the problem that is considered to be reliable. The second problem is finding efficient ways to represent this complexity of knowledge so that the program can utilize it efficiently. You also might want to include concrete problems to analyze for your central problem suggestions. That would help define the problem a bit better for analysis. Dave On Wed, Jul 14, 2010 at 8:30 AM, Jim Bromer jimbro...@gmail.com wrote: On Tue, Jul 13, 2010 at 9:05 PM, Jim Bromer jimbro...@gmail.com wrote: Even if you refined your model until it was just right, you would have only caught up to everyone else with a solution to a narrow AI problem. I did not mean that you would just have a solution to a narrow AI problem, but that your solution, if put in the form of scoring of points on the basis of the observation *of definitive* events, would constitute a narrow AI method. The central unsolved problem, in my view, is: How can hypotheses be conceptually integrated along with the observable definitive events of the problem to form good explanatory connections that can mesh well with other knowledge about the problem that is considered to be reliable. The second problem is finding efficient ways to represent this complexity of knowledge so that the program can utilize it efficiently. *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] How do we Score Hypotheses?
Actually, I just realized that there is a way to included inductive knowledge and experience into this algorithm. Inductive knowledge and experience about a specific object or object type can be exploited to know which hypotheses in the past were successful, and therefore which hypothesis is most likely. By choosing the most likely hypothesis first, we skip a lot of messy hypothesis comparison processing and analysis. If we choose the right hypothesis first, all we really have to do is verify that this hypothesis reveals in the data what we expect to be there. If we confirm what we expect, that is reason enough not to look for other hypotheses because the data is explained by what we originally believed to be likely. We only look for additional hypotheses when we find something unexplained. And even then, we don't look at the whole problem. We only look at what we have to to explain the unexplained data. In fact, we could even ignore the unexplained data if we believe, from experience, that it isn't pertinent. I discovered this because I'm analyzing how a series of hypotheses are navigated when analyzing images. It seems to me that it is done very similarly to way we do it. We sort of confirm what we expect and try to explain what we don't expect. We try out hypotheses in a sort of trial and error manor and see how each hypothesis affects what we find in the image. If we confirm things because of the hypothesis, we are likely to keep it. We keep going, navigating the tree of hypotheses, conflicts and unexpected observations until we find a good hypothesis. Something like that. I'm attempting to construct an algorithm for doing this as I analyze specific problems. Dave On Wed, Jul 14, 2010 at 10:22 AM, David Jones davidher...@gmail.com wrote: What do you mean by definitive events? I guess the first problem I see with my approach is that the movement of the window is also a hypothesis. I need to analyze it in more detail and see how the tree of hypotheses affects the hypotheses regarding the es on the windows. What I believe is that these problems can be broken down into types of hypotheses, types of events and types of relationships. then those types can be reasoned about in a general way. If possible, then you have a method for reasoning about any object that is covered by the types of hypotheses, events and relationships that you have defined. How to reason about specific objects should not be preprogrammed. But, I think the solution to this part of AGI is to find general ways to reason about a small set of concepts that can be combined to describe specific objects and situations. There are other parts to AGI that I am not considering yet. I believe the problem has to be broken down into separate pieces and understood before putting it back together into a complete system. I have not covered inductive learning for example, which would be an important part of AGI. I have also not yet incorporated learned experience into the algorithm, which is also important. The general AI problem is way too complicated to consider all at once. I simply can't solve hypothesis generation, comparison and disambiguation while at the same time solving induction and experience-based reasoning. It becomes unwieldly. So, I'm starting where I can and I'll work my way up to the full complexity of the problem. I don't really understand what you mean here: The central unsolved problem, in my view, is: How can hypotheses be conceptually integrated along with the observable definitive events of the problem to form good explanatory connections that can mesh well with other knowledge about the problem that is considered to be reliable. The second problem is finding efficient ways to represent this complexity of knowledge so that the program can utilize it efficiently. You also might want to include concrete problems to analyze for your central problem suggestions. That would help define the problem a bit better for analysis. Dave On Wed, Jul 14, 2010 at 8:30 AM, Jim Bromer jimbro...@gmail.com wrote: On Tue, Jul 13, 2010 at 9:05 PM, Jim Bromer jimbro...@gmail.com wrote: Even if you refined your model until it was just right, you would have only caught up to everyone else with a solution to a narrow AI problem. I did not mean that you would just have a solution to a narrow AI problem, but that your solution, if put in the form of scoring of points on the basis of the observation *of definitive* events, would constitute a narrow AI method. The central unsolved problem, in my view, is: How can hypotheses be conceptually integrated along with the observable definitive events of the problem to form good explanatory connections that can mesh well with other knowledge about the problem that is considered to be reliable. The second problem is finding efficient ways to represent this complexity of knowledge so that the program can utilize it efficiently. *agi
Re: [agi] Re: Huge Progress on the Core of AGI
Abram, Thanks for the clarification Abram. I don't have a single way to deal with uncertainty. I try not to decide on a method ahead of time because what I really want to do is analyze the problems and find a solution. But, at the same time. I have looked at the probabilistic approaches and they don't seem to be sufficient to solve the problem as they are now. So, my inclination is to use methods that don't gloss over important details. For me, the most important way of dealing with uncertainty is through explanatory-type reasoning. But, explanatory reasoning has not been well defined yet. So, the implementation is not yet clear. That's where I am now. I've begun to approach problems as follows. I try to break the problem down and answer the following questions: 1) How do we come up with or construct possible hypotheses. 2) How do we compare hypotheses to determine which is better. 3) How do we lower the uncertainty of hypotheses. 4) How do we determine the likelihood or strength of a single hypothesis all by itself. Is it sufficient on its own? With those questions in mind, the solution seems to be to break possible hypotheses down into pieces that are generally applicable. For example, in image analysis, a particular type of hypothesis might be related to 1) motion or 2) attachment relationships or 3) change or movement behavior of an object, etc. By breaking the possible hypotheses into very general pieces, you can apply them to just about any problem. With that as a tool, you can then develop general methods for resolving uncertainty of such hypotheses using explanatory scoring, consistency, and even statistical analysis. Does that make sense to you? Dave On Tue, Jul 13, 2010 at 2:29 AM, Abram Demski abramdem...@gmail.com wrote: PS-- I am not denying that statistics is applied probability theory. :) When I say they are different, what I mean is that saying I'm going to use probability theory and I'm going to use statistics tend to indicate very different approaches. Probability is a set of axioms, whereas statistics is a set of methods. The probability theory camp tends to be bayesian, whereas the stats camp tends to be frequentist. Your complaint that probability theory doesn't try to figure out why it was wrong in the 30% (or whatever) it misses is a common objection. Probability theory glosses over important detail, it encourages lazy thinking, etc. However, this all depends on the space of hypotheses being examined. Statistical methods will be prone to this objection because they are essentially narrow-AI methods: they don't *try* to search in the space of all hypotheses a human might consider. An AGI setup can and should have such a large hypothesis space. Note that AIXI is typically formulated as using a space of crisp (non-probabilistic) hypotheses, though probability theory is used to reason about them. This means no theory it considers will gloss over detail in this way: every theory completely explains the data. (I use AIXI as a convenient example, not because I agree with it.) --Abram --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
Mike, you are so full of it. There is a big difference between *can* and *don't*. You have no proof that programs can't handle anything you say that can't. On Tue, Jul 13, 2010 at 2:36 PM, Mike Tintner tint...@blueyonder.co.ukwrote: The first thing is to acknowledge that programs *don't* handle concepts - if you think they do, you must give examples. The reasons they can't, as presently conceived, is a) concepts encase a more or less *infinite diversity of forms* (even if only applying at first to a species of object) - *chair* for example as I've demonstrated embraces a vast open-ended diversity of radically different chair forms; higher order concepts like furniture embrace ... well, it's hard to think even of the parameters, let alone the diversity of forms, here. b) concepts are *polydomain*- not just multi- but open-endedly extensible in their domains; chair for example, can also refer to a person, skin in French, two humans forming a chair to carry s.o., a prize, etc. Basically concepts have a freeform realm or sphere of reference, and you can't have a setform, preprogrammed approach to defining that realm. There's no reason however why you can't mechanically and computationally begin to instantiate the kind of freeform approach I'm proposing. The most important obstacle is the setform mindset of AGI-ers - epitomised by Dave looking at squares, moving along set lines - setform objects in setform motion - when it would be more appropriate to look at something like snakes.- freeform objects in freeform motion. Concepts also - altho this is a huge subject - are *the* language of the general programs (as distinct from specialist programs, wh. is all we have right now) that must inform an AGI. Anyone proposing a grandscale AGI project like Ben's (wh. I def. wouldn't recommend) must crack the problem of conceptualisation more or less from the beginning. I'm not aware of anyone who has any remotely viable proposals here, are you? *From:* Jim Bromer jimbro...@gmail.com *Sent:* Tuesday, July 13, 2010 5:46 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI On Tue, Jul 13, 2010 at 10:07 AM, Mike Tintner tint...@blueyonder.co.ukwrote: And programs as we know them, don't and can't handle *concepts* - despite the misnomers of conceptual graphs/spaces etc wh are not concepts at all. They can't for example handle writing or shopping because these can only be expressed as flexible outlines/schemas as per ideograms. I disagree with this, and so this is proper focus for our disagreement. Although there are other aspects of the problem that we probably disagree on, this is such a fundamental issue, that nothing can get past it. Either programs can deal with flexible outlines/schema or they can't. If they can't then AGI is probably impossible. If they can, then AGI is probably possible. I think that we both agree that creativity and imagination is absolutely necessary aspects of intelligence. Jim Bromer *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
Mike, see below. On Tue, Jul 13, 2010 at 2:36 PM, Mike Tintner tint...@blueyonder.co.ukwrote: The first thing is to acknowledge that programs *don't* handle concepts - if you think they do, you must give examples. The reasons they can't, as presently conceived, is a) concepts encase a more or less *infinite diversity of forms* (even if only applying at first to a species of object) - *chair* for example as I've demonstrated embraces a vast open-ended diversity of radically different chair forms; higher order concepts like furniture embrace ... well, it's hard to think even of the parameters, let alone the diversity of forms, here. invoking infinity is insufficient argument to say that a program can't recognize an infinite number of forms. In fact, I can prove it. Lets say that all numbers are made of digits 0,1,2,3...9. If you can recognize just 9 digits, you can read infinitely large numbers. Another example, you can create an infinite number of very diverse shapes and forms out of clay. But, I can represent every last one of them using simple mesh models. The mesh models are made of a very small number of concepts: lines, points, distance constraints, etc. So, an infinite number of diverse concepts or forms can be modeled using a very small number of concepts. In conclusion, you have no proof at all that programs can't handle these things. You just THINK they can't. But, I for one, know you're dead wrong. b) concepts are *polydomain*- not just multi- but open-endedly extensible in their domains; chair for example, can also refer to a person, skin in French, two humans forming a chair to carry s.o., a prize, etc. A chair is defined by anything you can sit on. Anything you can sit on is defined by a certain type of form that you can actually learn inductively. It is not impossible to teach a computer to recognize things that could be sat on or even things that seem like they have the form of something that might be sat on. To say that a computer can never learn this is impossible for you to claim. You see, very diverse concepts can be represented by a small number of other concepts such as time, space, 3D form, etc. You claim is completely baseless. Basically concepts have a freeform realm or sphere of reference, and you can't have a setform, preprogrammed approach to defining that realm. you can if it covers base concepts which can represent larger concepts. There's no reason however why you can't mechanically and computationally begin to instantiate the kind of freeform approach I'm proposing. The most important obstacle is the setform mindset of AGI-ers - epitomised by Dave looking at squares, moving along set lines - setform objects in setform motion - when it would be more appropriate to look at something like snakes.- freeform objects in freeform motion. squares can move in an infinite number of ways. It is an experiment An exercise... to learn how AGI deals with uncertainty, even if the uncertainty is very limited. Clearly you have no imagination to understand why doing such experiments might be useful. You think moving squares is simple just because they are squares. But, you fail to realize that uncertainty can be generated out of even very simple systems. And so far you have never stated how you would deal with such uncertainty. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
Thanks Abram, I'll read up on it when I get a chance. On Tue, Jul 13, 2010 at 12:03 PM, Abram Demski abramdem...@gmail.comwrote: David, Yes, this makes sense to me. To go back to your original query, I still think you will find a rich community relevant to your work if you look into the MDL literature (which additionally does not rely on probability theory, though as I said I'd say it's equivalent). Perhaps this book might be helpful: http://www.amazon.com/Description-Principle-Adaptive-Computation-Learning/dp/0262072815/ref=sr_1_1?ie=UTF8s=booksqid=1279036776sr=8-1 It includes a (short-ish?) section comparing the pros/cons of MDL and Bayesianism, and examining some of the mathematical linkings between them-- with the aim of showing that MDL is a broader principle. I disagree there, of course. :) --Abram On Tue, Jul 13, 2010 at 10:01 AM, David Jones davidher...@gmail.comwrote: Abram, Thanks for the clarification Abram. I don't have a single way to deal with uncertainty. I try not to decide on a method ahead of time because what I really want to do is analyze the problems and find a solution. But, at the same time. I have looked at the probabilistic approaches and they don't seem to be sufficient to solve the problem as they are now. So, my inclination is to use methods that don't gloss over important details. For me, the most important way of dealing with uncertainty is through explanatory-type reasoning. But, explanatory reasoning has not been well defined yet. So, the implementation is not yet clear. That's where I am now. I've begun to approach problems as follows. I try to break the problem down and answer the following questions: 1) How do we come up with or construct possible hypotheses. 2) How do we compare hypotheses to determine which is better. 3) How do we lower the uncertainty of hypotheses. 4) How do we determine the likelihood or strength of a single hypothesis all by itself. Is it sufficient on its own? With those questions in mind, the solution seems to be to break possible hypotheses down into pieces that are generally applicable. For example, in image analysis, a particular type of hypothesis might be related to 1) motion or 2) attachment relationships or 3) change or movement behavior of an object, etc. By breaking the possible hypotheses into very general pieces, you can apply them to just about any problem. With that as a tool, you can then develop general methods for resolving uncertainty of such hypotheses using explanatory scoring, consistency, and even statistical analysis. Does that make sense to you? Dave On Tue, Jul 13, 2010 at 2:29 AM, Abram Demski abramdem...@gmail.comwrote: PS-- I am not denying that statistics is applied probability theory. :) When I say they are different, what I mean is that saying I'm going to use probability theory and I'm going to use statistics tend to indicate very different approaches. Probability is a set of axioms, whereas statistics is a set of methods. The probability theory camp tends to be bayesian, whereas the stats camp tends to be frequentist. Your complaint that probability theory doesn't try to figure out why it was wrong in the 30% (or whatever) it misses is a common objection. Probability theory glosses over important detail, it encourages lazy thinking, etc. However, this all depends on the space of hypotheses being examined. Statistical methods will be prone to this objection because they are essentially narrow-AI methods: they don't *try* to search in the space of all hypotheses a human might consider. An AGI setup can and should have such a large hypothesis space. Note that AIXI is typically formulated as using a space of crisp (non-probabilistic) hypotheses, though probability theory is used to reason about them. This means no theory it considers will gloss over detail in this way: every theory completely explains the data. (I use AIXI as a convenient example, not because I agree with it.) --Abram *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com -- Abram Demski http://lo-tho.blogspot.com/ http://groups.google.com/group/one-logic *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] How do we Score Hypotheses?
I've been trying to figure out how to score hypotheses. Do you guys have any constructive ideas about how to define the way you score hypotheses like these a little better? I'll define the problem below in detail. I know Abram mentioned MDL, which I'm about to look into. Does that even apply to this sort of thing? I came up with a hypothesis scoring idea. It goes as follows *Rule 1:* Hypotheses are compared only 1 at a time. *Rule 2:* If hypothesis 1 predicts/expects/anticipates something, then you add (+1) to its score and subtract (-1) from hypothesis 2 if it doesn't also anticipate the observation. (Note:When comparing only 2 hypotheses, it may actually not be necessary to subtract from the competing hypothesis I guess.) *Here is the specific problem I'm analyzing: *Let's say that you have two window objects that contain the same letter, such as the letter e. In frame 0, the first window object is visible. In frame 1, window 1 moves a bit. In frame 2 though, the second window object appears and completely occludes the first window object. So, if you only look at the letter e from frame 0 to frame 2, it looks like it never disappears and it just moves. But that's not what happens. There are two independent instances of the letter e. But, how do we get the algorithm to figure this out in a general way? How do we get it to compare the two possible hypotheses (1 object or two objects) and decide that one is better than the other? That is what the hypothesis scoring method is for. *Algorithm Description and Details* *Hypothesis 1:* there are two separate objects... there are two separate instances of the letter e *Hypothesis 2:* there is only one letter object... only one letter e that occurs in all the frames of the video. *Time 0: object 1* *Time 1: e moves rigidly with object 1* H1: +1 compared to h2 because we expect the e to move rigidly with the first object, rather than independently from the first object. H2: -1 compared to h1 because we don't expect the first object to move rigidly with e but h1 does. *Time 2: object 2 appears and completely occludes object 1. Object 1 and 2 both have the letter e on them. So, to a dumb algorithm, it looks as if the e moved between the two frames of the video.* H1: -1 compared to h2 because we don't expect what h2 expects. H2: +1 compared to h1 e moves independently of the first window *Time 3: e moves rigidly with object 2* H1: +1 compared to h2 e moves with second object. H2: -1 compared to h1 *Time 4: e moves rigidly with object 2* H1: +1 compared to h2 e moves with second object. H2: -1 compared to h1 *Time 5: e moves rigidly with object 2* H1: +1 compared to h2 e moves with second object. H2: -1 compared to h1 *After 5 video frames the score is: * H1: +3 H2: -3 Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
, find one of them that works with *unspecified kinds of actions and objects.* (Or you can always try and explain how formulae that are clearly designed to be setform can somehow simultaneously be freeform and embrace et cetera ). There are by the same token no branches of logic and maths that work with *unspecified kinds of actions and objects.* (Mathematicians who invent new formulae have to work with and develop new kinds of objects - but normal maths can't help them do this). The whole of rationality - incl. all rational technology - only works with specified kinds of actions and objects. **One of the most basic rationales of rationality is let's stop everyone farting around making their own versions of products with their own differently specified actions and objects; let's specify/standardise the actions and objects that everyone must use. Let's start making standard specification cherry cakes with standard ingredients, and standard mathematical sums with standard numbers and operations, and standard logical variables with standard meanings [and cut out any kind of et cetera] ** (And for much the same reason programs can't - aren't meant to - handle concepts. Every concept , like chair has to refer to a general class of objects embracing et ceteras - new, unspecified, yet-to-be-invented kinds of objects and ones that you simply haven't heard of yet, as well as specified, known kinds of object . Concepts are wonderful cognitive tools for embracing unspecified objects. Concepts, for example, like things, objects, actions, do something - anything all sorts of things - everything you can possibly think of even write totally new kinds of programs - anti-programs - et cetera - such concepts endow humans with massive creative freedom and scope of reference. You along with the whole of AI/AGI are effectively claiming that there is or can be a formula/program for dealing with the unknown - i.e. unknown kinds of objects. It's patent absurdity - and counter to the whole spirit of logic and rationality - in fact lunacy. You'll wonder in years to come how so smart people could be so dumb. Could think they're producing programs that can make anything - can make cars or cakes - any car or cake - when the rest of the world and his uncle can see that they're only producing very specific brands of car and cake (with very specific objects/parts). VW Beetles not cars let alone vehicles let alone forms of transportation let alone means of travel let alone universal programs. . I'm full of it? AI/AGI is full of the most amazing hype about its generality and creativity wh. you have clearly swallowed whole . Programs are simply specialist procedures for producing specialist products and procedures with specified kinds of actions and objects - they cannot deal with unspecified kinds of actions and objects, period. You won't produce any actual examples to the contrary. *From:* David Jones davidher...@gmail.com *Sent:* Tuesday, July 13, 2010 8:00 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI Correction: Mike, you are so full of it. There is a big difference between *can* and *don't*. You have no proof that programs can't handle anything you say [they] can't. On Tue, Jul 13, 2010 at 2:59 PM, David Jones davidher...@gmail.comwrote: Mike, you are so full of it. There is a big difference between *can* and *don't*. You have no proof that programs can't handle anything you say that can't. On Tue, Jul 13, 2010 at 2:36 PM, Mike Tintner tint...@blueyonder.co.ukwrote: The first thing is to acknowledge that programs *don't* handle concepts - if you think they do, you must give examples. The reasons they can't, as presently conceived, is a) concepts encase a more or less *infinite diversity of forms* (even if only applying at first to a species of object) - *chair* for example as I've demonstrated embraces a vast open-ended diversity of radically different chair forms; higher order concepts like furniture embrace ... well, it's hard to think even of the parameters, let alone the diversity of forms, here. b) concepts are *polydomain*- not just multi- but open-endedly extensible in their domains; chair for example, can also refer to a person, skin in French, two humans forming a chair to carry s.o., a prize, etc. Basically concepts have a freeform realm or sphere of reference, and you can't have a setform, preprogrammed approach to defining that realm. There's no reason however why you can't mechanically and computationally begin to instantiate the kind of freeform approach I'm proposing. The most important obstacle is the setform mindset of AGI-ers - epitomised by Dave looking at squares, moving along set lines - setform objects in setform motion - when it would be more appropriate to look at something like snakes.- freeform objects in freeform motion. Concepts also
Re: [agi] Re: Huge Progress on the Core of AGI
Thanks Abram, I know that probability is one approach. But there are many problems with using it in actual implementations. I know a lot of people will be angered by that statement and retort with all the successes that they have had using probability. But, the truth is that you can solve the problems many ways and every way has its pros and cons. I personally believe that probability has unacceptable cons if used all by itself. It must only be used when it is the best tool for the task. I do plan to use some probability within my approach. But only when it makes sense to do so. I do not believe in completely statistical solutions or completely Bayesian machine learning alone. A good example of when I might use it is when a particular hypothesis predicts something with 70% accuracy, well it may be better than any other hypothesis we can come up with so far. So, we may use that hypothesis. But, the 30% unexplained errors should be explained if possible with the resources and algorithms available, if at all possible. This is where my method differs from statistical methods. I want to build algorithms that resolve the 30% and explain it. For many problems, there are rules and knowledge that will solve them effectively. Probability should only be used when you cannot find a more accurate solution. Basically we should use probability when we don't know the factors involved, can't find any rules to explain the phenomena or we don't have the time and resources to figure it out. So you must simply guess at the most probable event without any rules for figuring out which event is more applicable under the current circumstances. So, in summary, probability definitely has its place. I just think that explanatory reasoning and other more accurate methods should be preferred whenever possible. Regarding learning the knowledge being the bigger problem, I completely agree. That is why I think it is so important to develop machine learning that can learn by direct observation of the environment. Without that, it is practically impossible to gather the knowledge required for AGI-type applications. We can learn this knowledge by analyzing the world automatically and generally through video. My step by step approach for learning and then applying the knowledge for agi is as follows: 1) Understand and learn about the environment(through Computer Vision for now and other sensory perceptions in the future) 2) learn about your own actions and how they affect the environment 3) learn about language and how it is associated with or related to the environment. 4) learn goals from language(such as through dedicated inputs). 5) Goal pursuit 6) Other Miscellaneous capabilities as needed Dave On Sat, Jul 10, 2010 at 8:40 PM, Abram Demski abramdem...@gmail.com wrote: David, Sorry for the slow response. I agree completely about expectations vs predictions, though I wouldn't use that terminology to make the distinction (since the two terms are near-synonyms in English, and I'm not aware of any technical definitions that are common in the literature). This is why I think probability theory is necessary: to formalize this idea of expectations. I also agree that it's good to utilize previous knowledge. However, I think existing AI research has tackled this over and over; learning that knowledge is the bigger problem. --Abram --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
Mike, Using the image itself as a template to match is possible. In fact it has been done before. But it suffers from several problems that would also need solving. 1) Images are 2D. I assume you are also referring to 2D outlines. Real objects are 3D. So, you're going to have to infer the shape of the object... which means you are no longer actually transforming the image itself. You are transforming a model of the image, which would have points, curves, dimensions, etc. Basically, a mathematical shape :) No matter how much you disapprove of encoding info, sometimes it makes sense to do it. 2) Creating the first outline and figuring out what to outline is not trivial at all. So, this method can only be used after you can do that. There is a lot more uncertainty involved here than you seem to realize. First, how do you even determine the outline? That is an unsolved problem. So, not only will you have to try many transformations with the right outline, you have to try many with wrong outlines, increase the possibilities (exponentially?). It looks like you need a way to score possibilities and decide which ones to try. 3) rock is a word and words are always learned by induction along with other types of reasoning before we can even consider it a type of object. So, you are starting with a somewhat unrepresentative or artificial problem. 4) Even the same rock can look very different from different perspectives. In fact, how do you even match the same rock? Please describe how your system would do this. It is not trivial at all. And you will soon see that there is an extremely large amount of uncertainty. Dealing with this type of uncertainty is the central problem of AGI. The central problem is not fluid schemas.Even if I used this method, I would be stuck with the same exact uncertainty problems. So, you're not going to get passed them like this. The same research on explanatory and non-monotonic type reasoning must still be done. 5) what is a fluid transform? You can't just throw out words. Please define it. You are going to realize that your representation is pretty much geometric anyway. Regardless, it will likely be equivalent. Are you going to try every possible transformation? Nope. That would be impossible. So, how do you decide what transformations to try? When is a transformation too large of a change to consider it the same rock? When is it too large to consider it a different rock? 6) Are you seriously going to transform every object you've every tried to outline? This is going to be prohibitively costly in terms of processing. Especially because you haven't defined how you're going to decide what to transform and what not to. So, before you can even use this algorithm, you're going to have to use something else to decide what is a possible candidate and what is not. On Fri, Jul 9, 2010 at 6:42 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Now let's see **you** answer a question. Tell me how any algorithmic/mathematical approach of any kind actual or in pure principle can be applied to recognize raindrops falling down a pane - and to predict their movement? Like I've said many times before, we can't predict everything, and we certainly shouldn't try. But http://www.pond5.com/stock-footage/263778/beautiful-rain-drops.html or to recognize a rock? http://www.handprint.com/HP/WCL/IMG/LPR/adams.jpg or a [filled] shopping bag? http://www.abc.net.au/reslib/200801/r215609_837743.jpg http://www.sustainableisgood.com/photos/uncategorized/2007/03/29/shoppingbags.jpg http://thegogreenblog.com/wp-content/uploads/2007/12/plastic_shopping_bag.jpg or if you want a real killer, google some vid clips of amoebas in oozing motion? PS In every case, I suggest, the brain observes different principles of transformation - for the most part unconsciously. And they can be of various kinds not just direct natural transformations, of course. It's possible, it occurs to me, that the principle that applies to rocks might just be something like whatever can be carved out of stone --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
I accidentally pressed something and it sent it early... this is a finished version: Mike, Using the image itself as a template to match is possible. In fact it has been done before. But it suffers from several problems that would also need solving. 1) Images are 2D. I assume you are also referring to 2D outlines. Real objects are 3D. So, you're going to have to infer the shape of the object... which means you are no longer actually transforming the image itself. You are transforming a model of the image, which would have points, curves, dimensions, etc. Basically, a mathematical shape :) No matter how much you disapprove of encoding info, sometimes it makes sense to do it. 2) Creating the first outline and figuring out what to outline is not trivial at all. So, this method can only be used after you can do that. There is a lot more uncertainty involved here than you seem to realize. First, how do you even determine the outline? That is an unsolved problem. So, not only will you have to try many transformations with the right outline, you have to try many with wrong outlines, increase the possibilities (exponentially?). It looks like you need a way to score possibilities and decide which ones to try. 3) rock is a word and words are always learned by induction along with other types of reasoning before we can even consider it a type of object. So, you are starting with a somewhat unrepresentative or artificial problem. 4) Even the same rock can look very different from different perspectives. In fact, how do you even match the same rock? Please describe how your system would do this. It is not trivial at all. And you will soon see that there is an extremely large amount of uncertainty. Dealing with this type of uncertainty is the central problem of AGI. The central problem is not fluid schemas.Even if I used this method, I would be stuck with the same exact uncertainty problems. So, you're not going to get passed them like this. The same research on explanatory and non-monotonic type reasoning must still be done. 5) what is a fluid transform? You can't just throw out words. Please define it. You are going to realize that your representation is pretty much geometric anyway. Regardless, it will likely be equivalent. Are you going to try every possible transformation? Nope. That would be impossible. So, how do you decide what transformations to try? When is a transformation too large of a change to consider it the same rock? When is it too large to consider it a different rock? 6) Are you seriously going to transform every object you've every tried to outline? This is going to be prohibitively costly in terms of processing. Especially because you haven't defined how you're going to decide what to transform and what not to. So, before you can even use this algorithm, you're going to have to use something else to decide what is a possible candidate and what is not. On Fri, Jul 9, 2010 at 6:42 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Now let's see **you** answer a question. Tell me how any algorithmic/mathematical approach of any kind actual or in pure principle can be applied to recognize raindrops falling down a pane - and to predict their movement? Like I've said many times before, we can't predict everything, and we certainly shouldn't try. But we should expect what might happen. Raindrops are probably recognized as an unexpected distortion when it occurs on a window. When its not on a window, it is still a sort of distortion of brightness and just a small object with different contrast. You're right that geometric definitions are not the right way to recognize that. It would have to use a different method to remember the features/properties of raindrops and how they appeared, such as the contrast, size, quantity, location, context, etc. http://www.pond5.com/stock-footage/263778/beautiful-rain-drops.html or to recognize a rock? A specific rock could be recognized with geometric definitions. Texture is certainly important, size, context (very important), etc. If we are talking about the category rock, that's different than the instance of a rock. The category of a rock probably needs a description of the types of properties that rocks have, such as the types of curves, texture, sizes, interactions, behavior, etc. Exactly how you do it, I haven't decided. I'm not at that point yet. http://www.handprint.com/HP/WCL/IMG/LPR/adams.jpg or a [filled] shopping bag? same as the rock. http://www.abc.net.au/reslib/200801/r215609_837743.jpg http://www.sustainableisgood.com/photos/uncategorized/2007/03/29/shoppingbags.jpg http://thegogreenblog.com/wp-content/uploads/2007/12/plastic_shopping_bag.jpg or if you want a real killer, google some vid clips of amoebas in oozing motion? same. PS In every case, I suggest, the brain observes different principles of transformation - for the most part unconsciously. And they can be of various kinds not just direct natural
Re: [agi] Re: Huge Progress on the Core of AGI
or reason. Here is a graphic demonstration of what you're trying to claim - in effect, you're saying geometry can define 'a piece of plasticine' [and by extension any standard transformation of a piece of plasticine as in a playroom] That's a nonsense. A piece of plasticine is a **freeform** object - it can be transformed into an unlimited diversity of shapes/forms (albeit with constraints). Formulae - the formulae of geometry - can only define **set form** objects, with a precise form and structure. There are no exceptions. Black is not white. Homogeneous is not heterogeneous. Set form is not freeform. All the objects I list - all irregular objects - are freeform objects. You are ironically misunderstanding the very foundations and rationale of geometry. Geometry - with its set form forms - was invented precisely because mathematicians didn't like the freeform nature of the world - wanted to create set forms (in the footsteps of the rational technologists who preceded them) - that they could control and reduce to formulae and reproduce with ease. Freeform rocks are a lot more complex to draw and make and reproduce than set form rectangular bricks. Set forms are not free forms. They are the opposite. (And while you and others will continue to *claim* in theory absolute setform=freeform nonsense, you will in practice always, always stick to setform objects. Some part of you knows the v.obvious truth ). *From:* David Jones davidher...@gmail.com *Sent:* Saturday, July 10, 2010 3:51 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI Mike, Using the image itself as a template to match is possible. In fact it has been done before. But it suffers from several problems that would also need solving. 1) Images are 2D. I assume you are also referring to 2D outlines. Real objects are 3D. So, you're going to have to infer the shape of the object... which means you are no longer actually transforming the image itself. You are transforming a model of the image, which would have points, curves, dimensions, etc. Basically, a mathematical shape :) No matter how much you disapprove of encoding info, sometimes it makes sense to do it. 2) Creating the first outline and figuring out what to outline is not trivial at all. So, this method can only be used after you can do that. There is a lot more uncertainty involved here than you seem to realize. First, how do you even determine the outline? That is an unsolved problem. So, not only will you have to try many transformations with the right outline, you have to try many with wrong outlines, increase the possibilities (exponentially?). It looks like you need a way to score possibilities and decide which ones to try. 3) rock is a word and words are always learned by induction along with other types of reasoning before we can even consider it a type of object. So, you are starting with a somewhat unrepresentative or artificial problem. 4) Even the same rock can look very different from different perspectives. In fact, how do you even match the same rock? Please describe how your system would do this. It is not trivial at all. And you will soon see that there is an extremely large amount of uncertainty. Dealing with this type of uncertainty is the central problem of AGI. The central problem is not fluid schemas.Even if I used this method, I would be stuck with the same exact uncertainty problems. So, you're not going to get passed them like this. The same research on explanatory and non-monotonic type reasoning must still be done. 5) what is a fluid transform? You can't just throw out words. Please define it. You are going to realize that your representation is pretty much geometric anyway. Regardless, it will likely be equivalent. Are you going to try every possible transformation? Nope. That would be impossible. So, how do you decide what transformations to try? When is a transformation too large of a change to consider it the same rock? When is it too large to consider it a different rock? 6) Are you seriously going to transform every object you've every tried to outline? This is going to be prohibitively costly in terms of processing. Especially because you haven't defined how you're going to decide what to transform and what not to. So, before you can even use this algorithm, you're going to have to use something else to decide what is a possible candidate and what is not. On Fri, Jul 9, 2010 at 6:42 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Now let's see **you** answer a question. Tell me how any algorithmic/mathematical approach of any kind actual or in pure principle can be applied to recognize raindrops falling down a pane - and to predict their movement? Like I've said many times before, we can't predict everything, and we certainly shouldn't try. But http://www.pond5.com/stock-footage/263778/beautiful-rain-drops.html
Re: [agi] Re: Huge Progress on the Core of AGI
On Sat, Jul 10, 2010 at 5:02 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Dave:You can't solve the problems with your approach either This is based on knowledge of what examples? Zero? It is based on the fact that you have refused to show how you deal with uncertainty. You haven't even conceded that there is uncertainty. I know for a fact that your method cannot solve the uncertainty, because it doesn't even consider that there might be any uncertainty. It is not a solution to anything. It is a mere suggestion of a way to compare objects. It isn't even a way to match them! So, when you're done comparing, your method only says it is different by this much. Well, what the hell does that do for you? Nothing at all. So, clearly my statement that your approach doesn't solve anything is well based. Yet, your claim that my approach is wrong is very poorly based. Your main disagreement is my simplification of the problem. That doesn't mean anything. I can go back and forth between the simple version and the more complex version whenever I want to after I've gained understanding through experiments on the simpler version. There is nothing wrong with the approach I am taking. It is completely necessary to study the nature of the problems and the principles that can solve the problems. I have given you one instance of s.o. [a technologist not a philosopher like me] who is if only in broad principle, trying to proceed in a non-encoding, analog-comparison direction. There must be others who are however crudely trying and considering what can be broadly classified as analog approaches. How much do you know, or have you even thought about such approaches? [Of course, computing doesn't have to be either/or analog-digital but can be both] the approaches are equivalent. I don't even say that my approach is digital. If I find a reason to use an analog approach, I'll use it. But so far, I can't find any reason to do so. BTW, you would be wiser to realize that analog can likely be well represented by digital encoding for the problems we are discussing. I see absolutely no reason to think analog is better than digital for any of these problems. You simply have a bias against my approach. And bias is not sufficient reason to disagree with me. My point 6) BTW is irrefutable, completely irrefutable, and puts a finger bang on why geometry obviously cannot cope with real objects, ( although I can and must, do a much more extensive job of exposition). That is ridiculous. First of all, a plastic bag can easily be represented geometrically as a mesh with length constraints and connectivity constraints. Of course it doesn't represent every possible transformation of the bag. It doesn't even make sense to store such a representation. In fact, its not possible. Your claim that geometry can't represent a plastic bag is downright dumb and trivially refutable. You could easily use your own ideas then to transform the mesh for matching, although I still claim this is not the right way to always match objects. In fact, I would dare say it is often the wrong way to match objects because of the processing and time cost. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
Mike, On Thu, Jul 8, 2010 at 6:52 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Isn't the first problem simply to differentiate the objects in a scene? Well, that is part of the movement problem. If you say something moved, you are also saying that the objects in the two or more video frames are the same instance. (Maybe the most important movement to begin with is not the movement of the object, but of the viewer changing their POV if only slightly - wh. won't be a factor if you're looking at a screen) Maybe, but this problem becomes kind of trivial in a 2D environment, assuming you don't allow rotation of the POV. Moving the POV would simply translate all the objects linearly. If you make it a 3D environment, it becomes significantly more complicated. I could work on 3D, which I will, but I'm not sure I should start there. I probably should consider it though and see what complications it adds to the problem and how they might be solved. And that I presume comes down to being able to put a crude, highly tentative, and fluid outline round them (something that won't be neces. if you're dealing with squares?) . Without knowing v. little if anything about what kind of objects they are. As an infant most likely does. {See infants' drawings and how they evolve v. gradually from a v. crude outline blob that at first can represent anything - that I'm suggesting is a replay of how visual perception developed). The fluid outline or image schema is arguably the basis of all intelligence - just about everything AGI is based on it. You need an outline for instance not just of objects, but of where you're going, and what you're going to try and do - if you want to survive in the real world. Schemas connect everything AGI. And it's not a matter of choice - first you have to have an outline/sense of the whole - whatever it is - before you can start filling in the parts. Well, this is the question. The solution is underdetermined, which means that a right solution is not possible to know with complete certainty. So, you may take the approach of using contours to match objects, but that is certainly not the only way to approach the problem. Yes, you have to use local features in the image to group pixels together in some way. I agree with you there. Is using contours the right way? Maybe, but not by itself. You have to define the problem a little better than just saying that we need to construct an outline. The real problem/question is this: How do you determine the uncertainty of a hypothesis, lower it and also determine how good a hypothesis is, especially in comparison to other hypotheses? So, in this case, we are trying to use an outline comparison to determine the best match hypotheses between objects. But, that doesn't define how you score alternative hypotheses. That also is certainly not the only way to do it. You could use the details within the outline too. In fact, in some situations, this would be required to disambiguate between the possible hypotheses. P.S. It would be mindblowingly foolish BTW to think you can do better than the way an infant learns to see - that's an awfully big visual section of the brain there, and it works. I'm not trying to do better than the human brain. I am trying to solve the same problems that the brain solves in a different way, sometimes better than the brain, sometimes worse, sometimes equivalently. What would be foolish is to assume the only way to duplicate general intelligence is to copy the human brain. By taking this approach, you are forced to reverse engineer and understand something that is extremely difficult to reverse engineer. In addition, a solution that using the brain's design may not be economically feasible. So, approaching the problem by copying the human brain has additional risks. You may end up figuring out how the brain works and not be able to use it. In addition might not end up with a good understanding of what other solutions might be possible. Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
On Fri, Jul 9, 2010 at 10:04 AM, Mike Tintner tint...@blueyonder.co.ukwrote: Couple of quick comments (I'm still thinking about all this - but I'm confident everything AGI links up here). A fluid schema is arguably by its v. nature a method - a trial and error, arguably universal method. It links vision to the hand or any effector. Handling objects also is based on fluid schemas - you put out a fluid adjustably-shaped hand to grasp things. And even if you don't have hands, like a worm, and must grasp things with your body, and must grasp the ground under which you move, then too you must use fluid body schemas/maps. All concepts - the basis of language and before language, all intelligence - are also almost certainly fluid schemas (and not as you suggested, patterns). fluid schemas is not an actual algorithm. It is not clear how to go about implementing such a design. Even so, when you get into the details of actually implementing it, you will find yourself faced with the exact same problems I'm trying to solve. So, lets say you take the first frame and generate an initial fluid schema. What if an object disappears? What if the object changes? What if the object moves a little or a lot? What if a large number of changes occur at once, like one new thing suddenly blocking a bunch of similar stuff that is behind it? How far does your fluid schema have to be distorted for the algorithm to realize that it needs a new schema and can't use the same old one? You can't just say that all objects are always present and just distort the schema. What if two similar objects appear or both move and one disappears? How does your schema handle this? Regardless of whether you talk about hypotheses or schemas, it is the SAME problem. You can't avoid the fact that the whole thing is underdetermined and you need a way to score and compare hypotheses. If you disagree, please define your schema algorithm a bit more specifically. Then we would be able to analyze its pros and cons better. All creative problemsolving begins from concepts of what you want to do (and not formulae or algorithms as in rational problemsolving). Any suggestion to the contrary will not, I suggest, bear the slightest serious examination. Sure. I would point out though that children do stuff just to learn in the beginning. A good example is our desire to play. Playing is a strategy by which children learn new things even though they don't have a need for those things yet. It motivates us to learn for the future and not for any pressing present needs. No matter how you look at it, you will need algorithms for general intelligence. To say otherwise makes zero sense. No algorithms, no design. No matter what design you come up with, I call that an algorithm. Algorithms don't have to be formulaic or narrow. Keep an open mind about the world algorithm, unless you can suggest a better term to describe general AI algorithms. **Fluid schemas/concepts/fluid outlines are attempts-to-grasp-things - gropings.** Point 2 : I'd relook at your assumptions in all your musings - my impression is they all assume, unwittingly, an *adult* POV - the view of s.o. who already knows how to see - as distinct from an infant who is just learning to see and get to grips with an extremely blurred world, (even more blurred and confusing, I wouldn't be surprised, than that Prakash video). You're unwittingly employing top down, fully-formed-intelligence assumptions even while overtly trying to produce a learning system - you're looking for what an adult wants to know, rather than what an infant starting-from-almost-no-knowledge-of-the-world wants to know. If you accept the point in any way, major philosophical rethinking is required. this point doesn't really define at all how the approach should be changed or what approach to take. So, it doesn't change the way I approach the problem. You would really have to be more specific. For example, you could say that the infant doesn't even know how to group pixels, so it has to automatically learn that. I would have to disagree with this approach because I can't think of any reasonable algorithms that could reasonably explore possibilities. It doesn't seem better to me to describe the problem even more generally to the point where you are learning how to learn. This is what Abram was suggesting. But, as I said to him, you need a way to suggest and search for possible learning methods and then compare them. There doesn't seem to be a way to do this effectively. And so, you shouldn't over generalize in this way. As I said in the initial email(this week), there is no such thing as perfectly general and a silver bullet for solving any problem. So, I believe that even infants are born expecting what the world will be like. They aren't able to learn about any world. They are optimized to configure their brains for this world. *From:* David Jones davidher...@gmail.com *Sent:* Friday, July 09
Re: [agi] Re: Huge Progress on the Core of AGI
Mike, Please outline your algorithm for fluid schemas though. It will be clear when you do that you are faced with the exact same uncertainty problems I am dealing with and trying to solve. The problems are completely equivalent. Yours is just a specific approach that is not sufficiently defined. You have to define how you deal with uncertainty when using fluid schemas or even how to approach the task of figuring it out. Until then, its not a solution to anything. Dave On Fri, Jul 9, 2010 at 10:59 AM, Mike Tintner tint...@blueyonder.co.ukwrote: If fluid schemas - speaking broadly - are what is needed, (and I'm pretty sure they are), it's n.g. trying for something else. You can't substitute a square approach for a fluid amoeba outline approach. (And you will certainly need exactly such an approach to recognize amoeba's). If it requires a new kind of machine, or a radically new kind of instruction set for computers, then that's what it requires - Stan Franklin, BTW, is one person who does recognize, and is trying to deal with this problem - might be worth checking up on him. This is partly BTW why my instinct is that it may be better to start with tasks for robot hands*, because it should be possible to get them to apply a relatively flexible and fluid grip/handshape and grope for and experiment with differently shaped objects And if you accept the broad philosophy I've been outlining, then it does make sense that evolution should have started with touch as a more primary sense, well before it got to vision. *Or perhaps it may prove better to start with robot snakes/bodies or somesuch. *From:* David Jones davidher...@gmail.com *Sent:* Friday, July 09, 2010 3:22 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI On Fri, Jul 9, 2010 at 10:04 AM, Mike Tintner tint...@blueyonder.co.ukwrote: Couple of quick comments (I'm still thinking about all this - but I'm confident everything AGI links up here). A fluid schema is arguably by its v. nature a method - a trial and error, arguably universal method. It links vision to the hand or any effector. Handling objects also is based on fluid schemas - you put out a fluid adjustably-shaped hand to grasp things. And even if you don't have hands, like a worm, and must grasp things with your body, and must grasp the ground under which you move, then too you must use fluid body schemas/maps. All concepts - the basis of language and before language, all intelligence - are also almost certainly fluid schemas (and not as you suggested, patterns). fluid schemas is not an actual algorithm. It is not clear how to go about implementing such a design. Even so, when you get into the details of actually implementing it, you will find yourself faced with the exact same problems I'm trying to solve. So, lets say you take the first frame and generate an initial fluid schema. What if an object disappears? What if the object changes? What if the object moves a little or a lot? What if a large number of changes occur at once, like one new thing suddenly blocking a bunch of similar stuff that is behind it? How far does your fluid schema have to be distorted for the algorithm to realize that it needs a new schema and can't use the same old one? You can't just say that all objects are always present and just distort the schema. What if two similar objects appear or both move and one disappears? How does your schema handle this? Regardless of whether you talk about hypotheses or schemas, it is the SAME problem. You can't avoid the fact that the whole thing is underdetermined and you need a way to score and compare hypotheses. If you disagree, please define your schema algorithm a bit more specifically. Then we would be able to analyze its pros and cons better. All creative problemsolving begins from concepts of what you want to do (and not formulae or algorithms as in rational problemsolving). Any suggestion to the contrary will not, I suggest, bear the slightest serious examination. Sure. I would point out though that children do stuff just to learn in the beginning. A good example is our desire to play. Playing is a strategy by which children learn new things even though they don't have a need for those things yet. It motivates us to learn for the future and not for any pressing present needs. No matter how you look at it, you will need algorithms for general intelligence. To say otherwise makes zero sense. No algorithms, no design. No matter what design you come up with, I call that an algorithm. Algorithms don't have to be formulaic or narrow. Keep an open mind about the world algorithm, unless you can suggest a better term to describe general AI algorithms. **Fluid schemas/concepts/fluid outlines are attempts-to-grasp-things - gropings.** Point 2 : I'd relook at your assumptions in all your musings - my impression is they all assume, unwittingly
Re: [agi] Solomonoff Induction is Not Universal and Probability is not Prediction
Although I haven't studied Solomonoff induction yet, although I plan to read up on it, I've realized that people seem to be making the same mistake I was. People are trying to find one silver bullet method of induction or learning that works for everything. I've begun to realize that its OK if something doesn't work for everything. As long as it works on a large enough subset of problems to be useful. If you can figure out how to construct justifiable methods of induction for enough problems that you need to solve, then that is sufficient for AGI. This is the same mistake I made and it was the point I was trying to make in the recent email I sent. I kept trying to come up with algorithms for doing things and I could always find a test case to break it. So, now I've begun to realize that it's ok if it breaks sometimes! The question is, can you define an algorithm that breaks gracefully and which can figure out what problems it can be applied to and what problems it should not be applied to. If you can do that, then you can solve the problems where it is applicable, and avoid the problems where it is not. This is perfectly OK! You don't have to find a silver bullet method of induction or inference that works for everything! Dave On Fri, Jul 9, 2010 at 10:49 AM, Ben Goertzel b...@goertzel.org wrote: To make this discussion more concrete, please look at http://www.vetta.org/documents/disSol.pdf Section 2.5 gives a simple version of the proof that Solomonoff induction is a powerful learning algorithm in principle, and Section 2.6 explains why it is not practically useful. What part of that paper do you think is wrong? thx ben On Fri, Jul 9, 2010 at 9:54 AM, Jim Bromer jimbro...@gmail.com wrote: On Fri, Jul 9, 2010 at 7:56 AM, Ben Goertzel b...@goertzel.org wrote: If you're going to argue against a mathematical theorem, your argument must be mathematical not verbal. Please explain one of 1) which step in the proof about Solomonoff induction's effectiveness you believe is in error 2) which of the assumptions of this proof you think is inapplicable to real intelligence [apart from the assumption of infinite or massive compute resources] Solomonoff Induction is not a provable Theorem, it is therefore a conjecture. It cannot be computed, it cannot be verified. There are many mathematical theorems that require the use of limits to prove them for example, and I accept those proofs. (Some people might not.) But there is no evidence that Solmonoff Induction would tend toward some limits. Now maybe the conjectured abstraction can be verified through some other means, but I have yet to see an adequate explanation of that in any terms. The idea that I have to answer your challenges using only the terms you specify is noise. Look at 2. What does that say about your Theorem. I am working on 1 but I just said: I haven't yet been able to find a way that could be used to prove that Solomonoff Induction does not do what Matt claims it does. Z What is not clear is that no one has objected to my characterization of the conjecture as I have been able to work it out for myself. It requires an infinite set of infinitely computed probabilities of each infinite string. If this characterization is correct, then Matt has been using the term string ambiguously. As a primary sample space: A particular string. And as a compound sample space: All the possible individual cases of the substring compounded into one. No one has yet to tell of his mathematical experiments of using a Turing simulator to see what a finite iteration of all possible programs of a given length would actually look like. I will finish this later. On Fri, Jul 9, 2010 at 7:49 AM, Jim Bromer jimbro...@gmail.com wrote: Abram, Solomoff Induction would produce poor predictions if it could be used to compute them. Solomonoff induction is a mathematical, not verbal, construct. Based on the most obvious mapping from the verbal terms you've used above into mathematical definitions in terms of which Solomonoff induction is constructed, the above statement of yours is FALSE. If you're going to argue against a mathematical theorem, your argument must be mathematical not verbal. Please explain one of 1) which step in the proof about Solomonoff induction's effectiveness you believe is in error 2) which of the assumptions of this proof you think is inapplicable to real intelligence [apart from the assumption of infinite or massive compute resources] Otherwise, your statement is in the same category as the statement by the protagonist of Dostoesvky's Notes from the Underground -- I admit that two times two makes four is an excellent thing, but if we are to give everything its due, two times two makes five is sometimes a very charming thing too. ;-) Secondly, since it cannot be computed it is useless. Third, it is not the sort of
Re: [agi] Solomonoff Induction is Not Universal and Probability is not Prediction
The same goes for inference. There is no silver bullet method that is completely general and can infer anything. There is no general inference method. Sometimes it works, sometimes it doesn't. That is the nature of the complex world we live in. My current theory is that the more we try to find a single silver bullet, the more we will just break against the fact that none exists. On Fri, Jul 9, 2010 at 11:35 AM, David Jones davidher...@gmail.com wrote: Although I haven't studied Solomonoff induction yet, although I plan to read up on it, I've realized that people seem to be making the same mistake I was. People are trying to find one silver bullet method of induction or learning that works for everything. I've begun to realize that its OK if something doesn't work for everything. As long as it works on a large enough subset of problems to be useful. If you can figure out how to construct justifiable methods of induction for enough problems that you need to solve, then that is sufficient for AGI. This is the same mistake I made and it was the point I was trying to make in the recent email I sent. I kept trying to come up with algorithms for doing things and I could always find a test case to break it. So, now I've begun to realize that it's ok if it breaks sometimes! The question is, can you define an algorithm that breaks gracefully and which can figure out what problems it can be applied to and what problems it should not be applied to. If you can do that, then you can solve the problems where it is applicable, and avoid the problems where it is not. This is perfectly OK! You don't have to find a silver bullet method of induction or inference that works for everything! Dave On Fri, Jul 9, 2010 at 10:49 AM, Ben Goertzel b...@goertzel.org wrote: To make this discussion more concrete, please look at http://www.vetta.org/documents/disSol.pdf Section 2.5 gives a simple version of the proof that Solomonoff induction is a powerful learning algorithm in principle, and Section 2.6 explains why it is not practically useful. What part of that paper do you think is wrong? thx ben On Fri, Jul 9, 2010 at 9:54 AM, Jim Bromer jimbro...@gmail.com wrote: On Fri, Jul 9, 2010 at 7:56 AM, Ben Goertzel b...@goertzel.org wrote: If you're going to argue against a mathematical theorem, your argument must be mathematical not verbal. Please explain one of 1) which step in the proof about Solomonoff induction's effectiveness you believe is in error 2) which of the assumptions of this proof you think is inapplicable to real intelligence [apart from the assumption of infinite or massive compute resources] Solomonoff Induction is not a provable Theorem, it is therefore a conjecture. It cannot be computed, it cannot be verified. There are many mathematical theorems that require the use of limits to prove them for example, and I accept those proofs. (Some people might not.) But there is no evidence that Solmonoff Induction would tend toward some limits. Now maybe the conjectured abstraction can be verified through some other means, but I have yet to see an adequate explanation of that in any terms. The idea that I have to answer your challenges using only the terms you specify is noise. Look at 2. What does that say about your Theorem. I am working on 1 but I just said: I haven't yet been able to find a way that could be used to prove that Solomonoff Induction does not do what Matt claims it does. Z What is not clear is that no one has objected to my characterization of the conjecture as I have been able to work it out for myself. It requires an infinite set of infinitely computed probabilities of each infinite string. If this characterization is correct, then Matt has been using the term string ambiguously. As a primary sample space: A particular string. And as a compound sample space: All the possible individual cases of the substring compounded into one. No one has yet to tell of his mathematical experiments of using a Turing simulator to see what a finite iteration of all possible programs of a given length would actually look like. I will finish this later. On Fri, Jul 9, 2010 at 7:49 AM, Jim Bromer jimbro...@gmail.comwrote: Abram, Solomoff Induction would produce poor predictions if it could be used to compute them. Solomonoff induction is a mathematical, not verbal, construct. Based on the most obvious mapping from the verbal terms you've used above into mathematical definitions in terms of which Solomonoff induction is constructed, the above statement of yours is FALSE. If you're going to argue against a mathematical theorem, your argument must be mathematical not verbal. Please explain one of 1) which step in the proof about Solomonoff induction's effectiveness you believe is in error 2) which of the assumptions of this proof you think is inapplicable to real
Re: [agi] Re: Huge Progress on the Core of AGI
The way I define algorithms encompasses just about any intelligently designed system. So, call it what you want. I really wish you would stop avoiding the word. But, fine. I'll play your word game... Define your system please. And justify why or how it handles uncertainty. You said overlay a hand to see if it fits. How do you define fits? The truth is that it will never fit perfectly, so how do you define a good fit and a bad one? You will find that you end up with the same exact problems I am working on. You keep avoiding the need to define the system of fluid schemas. You're avoiding it because it's not a solution to anything and you can't define it without realizing that your idea doesn't pan out. So, I dare you. Define your fluid schemas without revealing the fatal flaw in your reasoning. Dave On Fri, Jul 9, 2010 at 12:05 PM, Mike Tintner tint...@blueyonder.co.ukwrote: There isn't an algorithm. It's basically a matter of overlaying shapes to see if they fit - much as you put one hand against another to see if they fit - much as you can overlay a hand to see if it fits and is capable of grasping an object - except considerably more fluid/ rougher. There has to be some instruction generating the process, but it's not an algorithm. How can you have an algorithm for recognizing amoebas - or rocks or a drop of water? They are not patterned entities - or by extension reducible to algorithms. You don't need to think too much about internal visual processes - you can just look,at the external objects-to-be-classified , the objects that make up this world, and see this. Just as you can look at a set of diverse patterns and see that they too are not reducible to any single formula/pattern/algorithm. We're talking about the fundamental structure of the universe and its contents. If this is right and God is an artist before he is a mathematician, then it won't do any good screaming about it, you're going to have to invent a way to do art, so to speak, on computers . Or you can pretend that dealing with mathematical squares will somehow help here - but it hasn't and won't. Do you think that a creative process like creating http://www.apocalyptic-theories.com/gallery/lastjudge/bosch.jpg started with an algorithm? There are other ways of solving problems than algorithms - the person who created each algorithm in the first place certainly didn't have one. *From:* David Jones davidher...@gmail.com *Sent:* Friday, July 09, 2010 4:20 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI Mike, Please outline your algorithm for fluid schemas though. It will be clear when you do that you are faced with the exact same uncertainty problems I am dealing with and trying to solve. The problems are completely equivalent. Yours is just a specific approach that is not sufficiently defined. You have to define how you deal with uncertainty when using fluid schemas or even how to approach the task of figuring it out. Until then, its not a solution to anything. Dave On Fri, Jul 9, 2010 at 10:59 AM, Mike Tintner tint...@blueyonder.co.ukwrote: If fluid schemas - speaking broadly - are what is needed, (and I'm pretty sure they are), it's n.g. trying for something else. You can't substitute a square approach for a fluid amoeba outline approach. (And you will certainly need exactly such an approach to recognize amoeba's). If it requires a new kind of machine, or a radically new kind of instruction set for computers, then that's what it requires - Stan Franklin, BTW, is one person who does recognize, and is trying to deal with this problem - might be worth checking up on him. This is partly BTW why my instinct is that it may be better to start with tasks for robot hands*, because it should be possible to get them to apply a relatively flexible and fluid grip/handshape and grope for and experiment with differently shaped objects And if you accept the broad philosophy I've been outlining, then it does make sense that evolution should have started with touch as a more primary sense, well before it got to vision. *Or perhaps it may prove better to start with robot snakes/bodies or somesuch. *From:* David Jones davidher...@gmail.com *Sent:* Friday, July 09, 2010 3:22 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI On Fri, Jul 9, 2010 at 10:04 AM, Mike Tintner tint...@blueyonder.co.ukwrote: Couple of quick comments (I'm still thinking about all this - but I'm confident everything AGI links up here). A fluid schema is arguably by its v. nature a method - a trial and error, arguably universal method. It links vision to the hand or any effector. Handling objects also is based on fluid schemas - you put out a fluid adjustably-shaped hand to grasp things. And even if you don't have hands, like a worm, and must grasp things with your body, and must grasp the ground under
[agi] Re: Huge Progress on the Core of AGI
I've learned something really interesting today. I realized that general rules of inference probably don't really exists. There is no such thing as complete generality for these problems. The rules of inference that work for one environment would fail in alien environments. So, I have to modify my approach to solving these problems. As I studied over simplified problems, I realized that there are probably an infinite number of environments with their own behaviors that are not representative of the environments we want to put a general AI in. So, it is not ok to just come up with any case study and solve it. The case study has to actually be representative of a problem we want to solve in an environment we want to apply AI. Otherwise the solution required will take too long to develop because of it tries to accommodate too much generality. As I mentioned, such a general solution is likely impossible. So, someone could easily get stuck trying to solve an impossible task of creating one general solution to too many problems that don't allow a general solution. The best course is a balance between the time required to write a very general solution and the time required to write less general solutions for multiple problem types and environments. The best way to do this is to choose representative case studies to solve and make sure the solutions are truth-tropic and justified for the environments they are to be applied. Dave On Sun, Jun 27, 2010 at 1:31 AM, David Jones davidher...@gmail.com wrote: A method for comparing hypotheses in explanatory-based reasoning: * We prefer the hypothesis or explanation that ***expects* more observations. If both explanations expect the same observations, then the simpler of the two is preferred (because the unnecessary terms of the more complicated explanation do not add to the predictive power).* *Why are expected events so important?* They are a measure of 1) explanatory power and 2) predictive power. The more predictive and the more explanatory a hypothesis is, the more likely the hypothesis is when compared to a competing hypothesis. Here are two case studies I've been analyzing from sensory perception of simplified visual input: The goal of the case studies is to answer the following: How do you generate the most likely motion hypothesis in a way that is general and applicable to AGI? *Case Study 1)* Here is a link to an example: animated gif of two black squares move from left to righthttp://practicalai.org/images/CaseStudy1.gif. *Description: *Two black squares are moving in unison from left to right across a white screen. In each frame the black squares shift to the right so that square 1 steals square 2's original position and square two moves an equal distance to the right. *Case Study 2) *Here is a link to an example: the interrupted squarehttp://practicalai.org/images/CaseStudy2.gif. *Description:* A single square is moving from left to right. Suddenly in the third frame, a single black square is added in the middle of the expected path of the original black square. This second square just stays there. So, what happened? Did the square moving from left to right keep moving? Or did it stop and then another square suddenly appeared and moved from left to right? *Here is a simplified version of how we solve case study 1: *The important hypotheses to consider are: 1) the square from frame 1 of the video that has a very close position to the square from frame 2 should be matched (we hypothesize that they are the same square and that any difference in position is motion). So, what happens is that in each two frames of the video, we only match one square. The other square goes unmatched. 2) We do the same thing as in hypothesis #1, but this time we also match the remaining squares and hypothesize motion as follows: the first square jumps over the second square from left to right. We hypothesize that this happens over and over in each frame of the video. Square 2 stops and square 1 jumps over it over and over again. 3) We hypothesize that both squares move to the right in unison. This is the correct hypothesis. So, why should we prefer the correct hypothesis, #3 over the other two? Well, first of all, #3 is correct because it has the most explanatory power of the three and is the simplest of the three. Simpler is better because, with the given evidence and information, there is no reason to desire a more complicated hypothesis such as #2. So, the answer to the question is because explanation #3 expects the most observations, such as: 1) the consistent relative positions of the squares in each frame are expected. 2) It also expects their new positions in each from based on velocity calculations. 3) It expects both squares to occur in each frame. Explanation 1 ignores 1 square from each frame of the video, because it can't match it. Hypothesis #1 doesn't have a reason for why the a new square
[agi] Re: Huge Progress on the Core of AGI
An easy demonstration of this is visual illusions and even visual mistakes like one I sent to this list before. Our eyes sometimes infer things that are not true. It is absolutely necessary for such mistakes to occur because our sensory interpretation system is optimized for the world we expect to encounter, which didn't optical illusions during most of our development. A perfect solution to all visual problems and possible environments is [likely] impossible. It is ok to fail on optical illusions, since the failure has no fatal consequences, other than maybe thinking that there is a water puddle in the middle of the desert :). Dave On Thu, Jul 8, 2010 at 3:17 PM, David Jones davidher...@gmail.com wrote: I've learned something really interesting today. I realized that general rules of inference probably don't really exists. There is no such thing as complete generality for these problems. The rules of inference that work for one environment would fail in alien environments. So, I have to modify my approach to solving these problems. As I studied over simplified problems, I realized that there are probably an infinite number of environments with their own behaviors that are not representative of the environments we want to put a general AI in. So, it is not ok to just come up with any case study and solve it. The case study has to actually be representative of a problem we want to solve in an environment we want to apply AI. Otherwise the solution required will take too long to develop because of it tries to accommodate too much generality. As I mentioned, such a general solution is likely impossible. So, someone could easily get stuck trying to solve an impossible task of creating one general solution to too many problems that don't allow a general solution. The best course is a balance between the time required to write a very general solution and the time required to write less general solutions for multiple problem types and environments. The best way to do this is to choose representative case studies to solve and make sure the solutions are truth-tropic and justified for the environments they are to be applied. Dave On Sun, Jun 27, 2010 at 1:31 AM, David Jones davidher...@gmail.comwrote: A method for comparing hypotheses in explanatory-based reasoning: * We prefer the hypothesis or explanation that ***expects* more observations. If both explanations expect the same observations, then the simpler of the two is preferred (because the unnecessary terms of the more complicated explanation do not add to the predictive power).* *Why are expected events so important?* They are a measure of 1) explanatory power and 2) predictive power. The more predictive and the more explanatory a hypothesis is, the more likely the hypothesis is when compared to a competing hypothesis. Here are two case studies I've been analyzing from sensory perception of simplified visual input: The goal of the case studies is to answer the following: How do you generate the most likely motion hypothesis in a way that is general and applicable to AGI? *Case Study 1)* Here is a link to an example: animated gif of two black squares move from left to righthttp://practicalai.org/images/CaseStudy1.gif. *Description: *Two black squares are moving in unison from left to right across a white screen. In each frame the black squares shift to the right so that square 1 steals square 2's original position and square two moves an equal distance to the right. *Case Study 2) *Here is a link to an example: the interrupted squarehttp://practicalai.org/images/CaseStudy2.gif. *Description:* A single square is moving from left to right. Suddenly in the third frame, a single black square is added in the middle of the expected path of the original black square. This second square just stays there. So, what happened? Did the square moving from left to right keep moving? Or did it stop and then another square suddenly appeared and moved from left to right? *Here is a simplified version of how we solve case study 1: *The important hypotheses to consider are: 1) the square from frame 1 of the video that has a very close position to the square from frame 2 should be matched (we hypothesize that they are the same square and that any difference in position is motion). So, what happens is that in each two frames of the video, we only match one square. The other square goes unmatched. 2) We do the same thing as in hypothesis #1, but this time we also match the remaining squares and hypothesize motion as follows: the first square jumps over the second square from left to right. We hypothesize that this happens over and over in each frame of the video. Square 2 stops and square 1 jumps over it over and over again. 3) We hypothesize that both squares move to the right in unison. This is the correct hypothesis. So, why should we prefer the correct hypothesis, #3 over the other two
Re: [agi] Re: Huge Progress on the Core of AGI
It may not be possible to create a learning algorithm that can learn how to generally process images and other general AGI problems. This is for the same reason that completely general vision algorithms are likely impossible. I think that figuring out how to process sensory information intelligently requires either 1) impossible amounts of processing or 2) intelligent design and understanding by us. Maybe you could be more specific about how general learning algorithms would solve problems such as the one I'm tackling. But, I am extremely doubtful it can be done because the problems cannot be effectively described to such an algorithm. If you can't describe the problem, it can't search for solutions. If it can't search for solutions, you're basically stuck with evolution type algorithms, which require prohibitory amounts of processing. The reason that vision is so important for learning is that sensory perception is the foundation required to learn everything else. If you don't start with a foundational problem like this, you won't be representing the real nature of general intelligence problems that require extensive knowledge of the world to solve properly. Sensory perception is required to learn the information needed to understand everything else. Text and language for example, require extensive knowledge about the world to understand and especially to learn about. If you start with general learning algorithms on these unrepresentative problems, you will get stuck as we already have. So, it still makes a lot of sense to start with a concrete problem that does not require extensive amounts of previous knowledge to start learning. In fact, AGI requires that you not pre-program the AI with such extensive knowledge. So, lots of people are working on general learning algorithms that are unrepresentative of what is required for AGI because the algorithms don't have the knowledge needed to learn what they are trying to learn about. Regardless of how you look at it, my approach is definitely the right approach to AGI in my opinion. On Thu, Jul 8, 2010 at 5:02 PM, Abram Demski abramdem...@gmail.com wrote: David, That's why, imho, the rules need to be *learned* (and, when need be, unlearned). IE, what we need to work on is general learning algorithms, not general visual processing algorithms. As you say, there's not even such a thing as a general visual processing algorithm. Learning algorithms suffer similar environment-dependence, but (by their nature) not as severe... --Abram On Thu, Jul 8, 2010 at 3:17 PM, David Jones davidher...@gmail.com wrote: I've learned something really interesting today. I realized that general rules of inference probably don't really exists. There is no such thing as complete generality for these problems. The rules of inference that work for one environment would fail in alien environments. So, I have to modify my approach to solving these problems. As I studied over simplified problems, I realized that there are probably an infinite number of environments with their own behaviors that are not representative of the environments we want to put a general AI in. So, it is not ok to just come up with any case study and solve it. The case study has to actually be representative of a problem we want to solve in an environment we want to apply AI. Otherwise the solution required will take too long to develop because of it tries to accommodate too much generality. As I mentioned, such a general solution is likely impossible. So, someone could easily get stuck trying to solve an impossible task of creating one general solution to too many problems that don't allow a general solution. The best course is a balance between the time required to write a very general solution and the time required to write less general solutions for multiple problem types and environments. The best way to do this is to choose representative case studies to solve and make sure the solutions are truth-tropic and justified for the environments they are to be applied. Dave On Sun, Jun 27, 2010 at 1:31 AM, David Jones davidher...@gmail.comwrote: A method for comparing hypotheses in explanatory-based reasoning: * We prefer the hypothesis or explanation that ***expects* more observations. If both explanations expect the same observations, then the simpler of the two is preferred (because the unnecessary terms of the more complicated explanation do not add to the predictive power).* *Why are expected events so important?* They are a measure of 1) explanatory power and 2) predictive power. The more predictive and the more explanatory a hypothesis is, the more likely the hypothesis is when compared to a competing hypothesis. Here are two case studies I've been analyzing from sensory perception of simplified visual input: The goal of the case studies is to answer the following: How do you generate the most likely motion hypothesis in a way that is general
Re: [agi] Re: Huge Progress on the Core of AGI
Abram, Yeah, I would have to object for a couple reasons. First, prediction requires previous knowledge. So, even if you make that your primary goal, you're still going to have my research goals as the prerequisite: which are to process visual information in a more general way and learn about the environment in a more general way. Second, not everything is predictable. Certainly, we should not try to predict everything. Only after we have experience, can we actually predict anything. Even then, it's not precise prediction, like predicting the next frame of a video. It's more like having knowledge of what is quite likely to occur, or maybe an approximate prediction, but not guaranteed in the least. For example, based on previous experience, striking a match will light it. But, sometimes it doesn't light, and that too is expected to occur sometimes. We definitely don't predict the next image we'll see when it lights though. We just have expectations for what we might see and this helps us interpret the image effectively. We should try to expect certain outcomes or possible outcomes though. You could call that prediction, but it's not quite the same. The things we are more likely to see should be attempted as an explanation first and preferred if not given a reason to think otherwise. Dave On Thu, Jul 8, 2010 at 5:51 PM, Abram Demski abramdem...@gmail.com wrote: David, How I'd present the problem would be predict the next frame, or more generally predict a specified portion of video given a different portion. Do you object to this approach? --Abram On Thu, Jul 8, 2010 at 5:30 PM, David Jones davidher...@gmail.com wrote: It may not be possible to create a learning algorithm that can learn how to generally process images and other general AGI problems. This is for the same reason that completely general vision algorithms are likely impossible. I think that figuring out how to process sensory information intelligently requires either 1) impossible amounts of processing or 2) intelligent design and understanding by us. Maybe you could be more specific about how general learning algorithms would solve problems such as the one I'm tackling. But, I am extremely doubtful it can be done because the problems cannot be effectively described to such an algorithm. If you can't describe the problem, it can't search for solutions. If it can't search for solutions, you're basically stuck with evolution type algorithms, which require prohibitory amounts of processing. The reason that vision is so important for learning is that sensory perception is the foundation required to learn everything else. If you don't start with a foundational problem like this, you won't be representing the real nature of general intelligence problems that require extensive knowledge of the world to solve properly. Sensory perception is required to learn the information needed to understand everything else. Text and language for example, require extensive knowledge about the world to understand and especially to learn about. If you start with general learning algorithms on these unrepresentative problems, you will get stuck as we already have. So, it still makes a lot of sense to start with a concrete problem that does not require extensive amounts of previous knowledge to start learning. In fact, AGI requires that you not pre-program the AI with such extensive knowledge. So, lots of people are working on general learning algorithms that are unrepresentative of what is required for AGI because the algorithms don't have the knowledge needed to learn what they are trying to learn about. Regardless of how you look at it, my approach is definitely the right approach to AGI in my opinion. On Thu, Jul 8, 2010 at 5:02 PM, Abram Demski abramdem...@gmail.comwrote: David, That's why, imho, the rules need to be *learned* (and, when need be, unlearned). IE, what we need to work on is general learning algorithms, not general visual processing algorithms. As you say, there's not even such a thing as a general visual processing algorithm. Learning algorithms suffer similar environment-dependence, but (by their nature) not as severe... --Abram On Thu, Jul 8, 2010 at 3:17 PM, David Jones davidher...@gmail.comwrote: I've learned something really interesting today. I realized that general rules of inference probably don't really exists. There is no such thing as complete generality for these problems. The rules of inference that work for one environment would fail in alien environments. So, I have to modify my approach to solving these problems. As I studied over simplified problems, I realized that there are probably an infinite number of environments with their own behaviors that are not representative of the environments we want to put a general AI in. So, it is not ok to just come up with any case study and solve it. The case study has to actually be representative of a problem we
Re: [agi] Open Sets vs Closed Sets
narrow AI is a term that describes the solution to a problem, not the problem. It is a solution with a narrow scope. General AI on the other hand should have a much larger scope than narrow ai and be able to handle unforseen circumstances. What I don't think you realize is that open sets can be described by closed sets. Here is an example from my own research. The set of objects I'm allowing in the simplest case studies so far are black squares. This is a closed set. But, the number, movement and relative positions of these squares is an open set. I can define an infinite number of ways in which a 0 to infinite number of black squares can move. If I define a general AI algorithm, it should be able to handle the infinite subset of the open set that is representative of some aspect of the real world. We could also study case studies that are not representative of the environment though. The example I just gave is a completely open set, yet an algorithm could handle such an open set, and I am designing for it. So, your claim that no one is studying or handling such things is not right. Dave On Wed, Jun 30, 2010 at 8:58 AM, Mike Tintner tint...@blueyonder.co.ukwrote: I'd like opinions on terminology here. IMO the opposition of closed sets vs open sets is fundamental to the difference between narrow AI and AGI. However I notice that these terms have different meanings to mine in maths. What I mean is: closed set: contains a definable number and *kinds/species* of objects open set: contains an undefinable number and *kinds/species* of objects (what we in casual, careless conversation describe as containing all kinds of things); the rules of an open set allow adding new kinds of things ad infinitum Narrow AI's operate in artificial environments containing closed sets of objects - all of wh. are definable. AGI's operate in real world environments containing open sets of objects - some of wh. will be definable, and some definitely not To engage in any real world activity, like walking down a street or searching/tidying a room or reading a science book/text is to operate with open sets of objects, because the next field of operations - the next street or room or text - may and almost certainly will have unpredictably different kinds of objects from the last. Any objections to my use of these terms, or suggestions that I should use others? *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com/ --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Huge Progress on the Core of AGI
Nice Occam's Razor argument. I understood it simply because I knew there are always an infinite number of possible explanations for every observation that are more complicated than the simplest explanation. So, without a reason to choose one of those other interpretations, then why choose it? You could look for reasons in complex environments, but it would likely be more efficient to wait for a reason to need a better explanation. It's more efficient to wait for an inconsistency than to search an infinite set without a reason to do so. Dave On Fri, Jul 2, 2010 at 6:08 PM, Matt Mahoney matmaho...@yahoo.com wrote: Jim, to address all of your points, Solomonoff induction claims that the probability of a string is proportional to the number of programs that output the string, where each program M is weighted by 2^-|M|. The probability is dominated by the shortest program (Kolmogorov complexity), but it is not exactly the same. The difference is small enough that we may neglect it, just as we neglect differences that depend on choice of language. Here is the proof that Kolmogorov complexity is not computable. Suppose it were. Then I could test the Kolmogorov complexity of strings in increasing order of length (breaking ties lexicographically) and describe the first string that cannot be described in less than a million bits, contradicting the fact that I just did. (Formally, I could write a program that outputs the first string whose Kolmogorov complexity is at least n bits, choosing n to be larger than my program). Here is the argument that Occam's Razor and Solomonoff distribution must be true. Consider all possible probability distributions p(x) over any infinite set X of possible finite strings x, i.e. any X = {x: p(x) 0} that is infinite. All such distributions must favor shorter strings over longer ones. Consider any x in X. Then p(x) 0. There can be at most a finite number (less than 1/p(x)) of strings that are more likely than x, and therefore an infinite number of strings which are less likely than x. Of this infinite set, only a finite number (less than 2^|x|) can be shorter than x, and therefore there must be an infinite number that are longer than x. So for each x we can partition X into 4 subsets as follows: - shorter and more likely than x: finite - shorter and less likely than x: finite - longer and more likely than x: finite - longer and less likely than x: infinite. So in this sense, any distribution over the set of strings must favor shorter strings over longer ones. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* Jim Bromer jimbro...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Fri, July 2, 2010 4:09:38 PM *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI On Fri, Jul 2, 2010 at 2:25 PM, Jim Bromer jimbro...@gmail.com wrote: There cannot be a one to one correspondence to the representation of the shortest program to produce a string and the strings that they produce. This means that if the consideration of the hypotheses were to be put into general mathematical form it must include the potential of many to one relations between candidate programs (or subprograms) and output strings. But, there is also no way to determine what the shortest program is, since there may be different programs that are the same length. That means that there is a many to one relation between programs and program length. So the claim that you could just iterate through programs *by length* is false. This is the goal of algorithmic information theory not a premise of a methodology that can be used. So you have the diagonalization problem. A counter argument is that there are only a finite number of Turing Machine programs of a given length. However, since you guys have specifically designated that this theorem applies to any construction of a Turing Machine it is not clear that this counter argument can be used. And there is still the specific problem that you might want to try a program that writes a longer program to output a string (or many strings). Or you might want to write a program that can be called to write longer programs on a dynamic basis. I think these cases, where you might consider a program that outputs a longer program, (or another instruction string for another Turing Machine) constitutes a serious problem, that at the least, deserves to be answered with sound analysis. Part of my original intuitive argument, that I formed some years ago, was that without a heavy constraint on the instructions for the program, it will be practically impossible to test or declare that some program is indeed the shortest program. However, I can't quite get to the point now that I can say that there is definitely a diagonalization problem. Jim Bromer *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ |
[agi] Re: Huge Progress on the Core of AGI
If anyone has any knowledge of or references to the state of the art in explanation-based reasoning, can you send me keywords or links? I've read some through google, but I'm not really satisfied with anything I've found. Thanks, Dave On Sun, Jun 27, 2010 at 1:31 AM, David Jones davidher...@gmail.com wrote: A method for comparing hypotheses in explanatory-based reasoning: * We prefer the hypothesis or explanation that ***expects* more observations. If both explanations expect the same observations, then the simpler of the two is preferred (because the unnecessary terms of the more complicated explanation do not add to the predictive power).* *Why are expected events so important?* They are a measure of 1) explanatory power and 2) predictive power. The more predictive and the more explanatory a hypothesis is, the more likely the hypothesis is when compared to a competing hypothesis. Here are two case studies I've been analyzing from sensory perception of simplified visual input: The goal of the case studies is to answer the following: How do you generate the most likely motion hypothesis in a way that is general and applicable to AGI? *Case Study 1)* Here is a link to an example: animated gif of two black squares move from left to righthttp://practicalai.org/images/CaseStudy1.gif. *Description: *Two black squares are moving in unison from left to right across a white screen. In each frame the black squares shift to the right so that square 1 steals square 2's original position and square two moves an equal distance to the right. *Case Study 2) *Here is a link to an example: the interrupted squarehttp://practicalai.org/images/CaseStudy2.gif. *Description:* A single square is moving from left to right. Suddenly in the third frame, a single black square is added in the middle of the expected path of the original black square. This second square just stays there. So, what happened? Did the square moving from left to right keep moving? Or did it stop and then another square suddenly appeared and moved from left to right? *Here is a simplified version of how we solve case study 1: *The important hypotheses to consider are: 1) the square from frame 1 of the video that has a very close position to the square from frame 2 should be matched (we hypothesize that they are the same square and that any difference in position is motion). So, what happens is that in each two frames of the video, we only match one square. The other square goes unmatched. 2) We do the same thing as in hypothesis #1, but this time we also match the remaining squares and hypothesize motion as follows: the first square jumps over the second square from left to right. We hypothesize that this happens over and over in each frame of the video. Square 2 stops and square 1 jumps over it over and over again. 3) We hypothesize that both squares move to the right in unison. This is the correct hypothesis. So, why should we prefer the correct hypothesis, #3 over the other two? Well, first of all, #3 is correct because it has the most explanatory power of the three and is the simplest of the three. Simpler is better because, with the given evidence and information, there is no reason to desire a more complicated hypothesis such as #2. So, the answer to the question is because explanation #3 expects the most observations, such as: 1) the consistent relative positions of the squares in each frame are expected. 2) It also expects their new positions in each from based on velocity calculations. 3) It expects both squares to occur in each frame. Explanation 1 ignores 1 square from each frame of the video, because it can't match it. Hypothesis #1 doesn't have a reason for why the a new square appears in each frame and why one disappears. It doesn't expect these observations. In fact, explanation 1 doesn't expect anything that happens because something new happens in each frame, which doesn't give it a chance to confirm its hypotheses in subsequent frames. The power of this method is immediately clear. It is general and it solves the problem very cleanly. *Here is a simplified version of how we solve case study 2:* We expect the original square to move at a similar velocity from left to right because we hypothesized that it did move from left to right and we calculated its velocity. If this expectation is confirmed, then it is more likely than saying that the square suddenly stopped and another started moving. Such a change would be unexpected and such a conclusion would be unjustifiable. I also believe that explanations which generate fewer incorrect expectations should be preferred over those that more incorrect expectations. The idea I came up with earlier this month regarding high frame rates to reduce uncertainty is still applicable. It is important that all generated hypotheses have as low uncertainty as possible given our constraints and resources available
Re: [agi] Re: Huge Progress on the Core of AGI
Thanks Matt, Right. But Occam's Razor is not complete. It says simpler is better, but 1) this only applies when two hypotheses have the same explanatory power and 2) what defines simpler? So, maybe what I want to know from the state of the art in research is: 1) how precisely do other people define simpler and 2) More importantly, how do you compare competing explanations/hypotheses that have more or less explanatory power. Simpler does not apply unless you are comparing equally explanatory hypotheses. For example, the simplest hypothesis for all visual interpretation is that everything in the first image is gone in the second image, and everything in the second image is a new object. Simple. Done. Solved :) right? Well, clearly a more complicated explanation is warranted because a more complicated explanation is more *explanatory* and a better explanation. So, why is it better? Can it be defined as better in a precise way so that you can compare arbitrary hypotheses or explanations? That is what I'm trying to learn about. I don't think much progress has been made in this area, but I'd like to know what other people have done and any successes they've had. Dave On Tue, Jun 29, 2010 at 10:29 AM, Matt Mahoney matmaho...@yahoo.com wrote: David Jones wrote: If anyone has any knowledge of or references to the state of the art in explanation-based reasoning, can you send me keywords or links? The simplest explanation of the past is the best predictor of the future. http://en.wikipedia.org/wiki/Occam's_razorhttp://en.wikipedia.org/wiki/Occam%27s_razor http://en.wikipedia.org/wiki/Occam%27s_razor http://www.scholarpedia.org/article/Algorithmic_probability http://www.scholarpedia.org/article/Algorithmic_probability -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Tue, June 29, 2010 9:05:45 AM *Subject:* [agi] Re: Huge Progress on the Core of AGI If anyone has any knowledge of or references to the state of the art in explanation-based reasoning, can you send me keywords or links? I've read some through google, but I'm not really satisfied with anything I've found. Thanks, Dave On Sun, Jun 27, 2010 at 1:31 AM, David Jones davidher...@gmail.comwrote: A method for comparing hypotheses in explanatory-based reasoning: * We prefer the hypothesis or explanation that ***expects* more observations. If both explanations expect the same observations, then the simpler of the two is preferred (because the unnecessary terms of the more complicated explanation do not add to the predictive power).* *Why are expected events so important?* They are a measure of 1) explanatory power and 2) predictive power. The more predictive and the more explanatory a hypothesis is, the more likely the hypothesis is when compared to a competing hypothesis. Here are two case studies I've been analyzing from sensory perception of simplified visual input: The goal of the case studies is to answer the following: How do you generate the most likely motion hypothesis in a way that is general and applicable to AGI? *Case Study 1)* Here is a link to an example: animated gif of two black squares move from left to righthttp://practicalai.org/images/CaseStudy1.gif. *Description: *Two black squares are moving in unison from left to right across a white screen. In each frame the black squares shift to the right so that square 1 steals square 2's original position and square two moves an equal distance to the right. *Case Study 2) *Here is a link to an example: the interrupted squarehttp://practicalai.org/images/CaseStudy2.gif. *Description:* A single square is moving from left to right. Suddenly in the third frame, a single black square is added in the middle of the expected path of the original black square. This second square just stays there. So, what happened? Did the square moving from left to right keep moving? Or did it stop and then another square suddenly appeared and moved from left to right? *Here is a simplified version of how we solve case study 1: *The important hypotheses to consider are: 1) the square from frame 1 of the video that has a very close position to the square from frame 2 should be matched (we hypothesize that they are the same square and that any difference in position is motion). So, what happens is that in each two frames of the video, we only match one square. The other square goes unmatched. 2) We do the same thing as in hypothesis #1, but this time we also match the remaining squares and hypothesize motion as follows: the first square jumps over the second square from left to right. We hypothesize that this happens over and over in each frame of the video. Square 2 stops and square 1 jumps over it over and over again. 3) We hypothesize that both squares move to the right in unison. This is the correct hypothesis. So, why should we prefer
Re: [agi] A Primary Distinction for an AGI
Mike, THIS is the flawed reasoning that causes people to ignore vision as the right way to create AGI. And I've finally come up with a great way to show you how wrong this reasoning is. I'll give you an extremely obvious argument that proves that vision requires much less knowledge to interpret than language does. Let's say that you have never been to egypt, you have never seen some particular movie before. But if you see the movie, an alien landscape, an alien world, a new place or any such new visual experience, you can immediately interpret it in terms of spacial, temporal, compositional and other relationships. Now, go to egypt and listen to them speak. Can you interpret it? Nope. Why?! Because you don't have enough information. The language itself does not contain any information to help you interpret it. We do not learn language simply by listening. We learn based on evidence from how the language is used and how it occurs in our daily lives. Without that experience, you cannot interpret it. But with vision, you do not need extra knowledge to interpret a new situation. You can recognize completely new objects without any training except for simply observing them in their natural state. I wish people understood this better. Dave On Tue, Jun 29, 2010 at 12:51 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Just off the cuff here - isn't the same true for vision? You can't learn vision from vision. Just as all NLP has no connection with the real world, and totally relies on the human programmer's knowledge of that world. Your visual program actually relies totally on your visual vocabulary - not its own. That is the inevitable penalty of processing unreal signals on a computer screen which are not in fact connected to the real world any more than the verbal/letter signals involved in NLP are. What you need to do - what anyone in your situation with anything like your asprations needs to do - is to hook up with a roboticist. Everyone here should be doing that. *From:* David Jones davidher...@gmail.com *Sent:* Tuesday, June 29, 2010 5:27 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] A Primary Distinction for an AGI You can't learn language from language without embedding way more knowledge than is reasonable. Language does not contain the information required for its interpretation. There is no *reason* to interpret the language into any of the infinite possible interpretaions. There is nothing to explain but it requires explanatory reasoning to determine the correct real world interpretation On Jun 29, 2010 10:58 AM, Matt Mahoney matmaho...@yahoo.com wrote: David Jones wrote: Natural language requires more than the words on the page in the real world. Of... Any knowledge that can be demonstrated over a text-only channel (as in the Turing test) can also be learned over a text-only channel. Cyc also is trying to store knowledge about a super complicated world in simplistic forms and al... Cyc failed because it lacks natural language. The vast knowledge store of the internet is unintelligible to Cyc. The average person can't use it because they don't speak Cycl and because they have neither the ability nor the patience to translate their implicit thoughts into augmented first order logic. Cyc's approach was understandable when they started in 1984 when they had neither the internet nor the vast computing power that is required to learn natural language from unlabeled examples like children do. Vision and other sensory interpretaion, on the other hand, do not require more info because that... Without natural language, your system will fail too. You don't have enough computing power to learn language, much less the million times more computing power you need to learn to see. -- Matt Mahoney, matmaho...@yahoo.com From: David Jones davidher...@gmail.com To: agi a...@v2.listbox.c... *Sent:* Mon, June 28, 2010 9:28:57 PM Subject: Re: [agi] A Primary Distinction for an AGI Natural language requires more than the words on the page in the real world. Of course that didn't ... *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com
Re: [agi] A Primary Distinction for an AGI
The point I was trying to make is that an approach that tries to interpret language just using language itself and without sufficient information or the means to realistically acquire that information, *should* fail. On the other hand, an approach that tries to interpret vision with minimal upfront knowledge needs *should* succeed because the knowledge required to automatically learn to interpret images is amenable to preprogramming. In addition, such knowledge must be pre-programmed. The knowledge for interpreting language though should not be pre-programmed. Dave On Tue, Jun 29, 2010 at 2:51 PM, Matt Mahoney matmaho...@yahoo.com wrote: David Jones wrote: I wish people understood this better. For example, animals can be intelligent even though they lack language because they can see. True, but an AGI with language skills is more useful than one without. And yes, I realize that language, vision, motor skills, hearing, and all the other senses and outputs are tied together. Skills in any area make learning the others easier. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Tue, June 29, 2010 1:42:51 PM *Subject:* Re: [agi] A Primary Distinction for an AGI Mike, THIS is the flawed reasoning that causes people to ignore vision as the right way to create AGI. And I've finally come up with a great way to show you how wrong this reasoning is. I'll give you an extremely obvious argument that proves that vision requires much less knowledge to interpret than language does. Let's say that you have never been to egypt, you have never seen some particular movie before. But if you see the movie, an alien landscape, an alien world, a new place or any such new visual experience, you can immediately interpret it in terms of spacial, temporal, compositional and other relationships. Now, go to egypt and listen to them speak. Can you interpret it? Nope. Why?! Because you don't have enough information. The language itself does not contain any information to help you interpret it. We do not learn language simply by listening. We learn based on evidence from how the language is used and how it occurs in our daily lives. Without that experience, you cannot interpret it. But with vision, you do not need extra knowledge to interpret a new situation. You can recognize completely new objects without any training except for simply observing them in their natural state. I wish people understood this better. Dave On Tue, Jun 29, 2010 at 12:51 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Just off the cuff here - isn't the same true for vision? You can't learn vision from vision. Just as all NLP has no connection with the real world, and totally relies on the human programmer's knowledge of that world. Your visual program actually relies totally on your visual vocabulary - not its own. That is the inevitable penalty of processing unreal signals on a computer screen which are not in fact connected to the real world any more than the verbal/letter signals involved in NLP are. What you need to do - what anyone in your situation with anything like your asprations needs to do - is to hook up with a roboticist. Everyone here should be doing that. *From:* David Jones davidher...@gmail.com *Sent:* Tuesday, June 29, 2010 5:27 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] A Primary Distinction for an AGI You can't learn language from language without embedding way more knowledge than is reasonable. Language does not contain the information required for its interpretation. There is no *reason* to interpret the language into any of the infinite possible interpretaions. There is nothing to explain but it requires explanatory reasoning to determine the correct real world interpretation On Jun 29, 2010 10:58 AM, Matt Mahoney matmaho...@yahoo.com wrote: David Jones wrote: Natural language requires more than the words on the page in the real world. Of... Any knowledge that can be demonstrated over a text-only channel (as in the Turing test) can also be learned over a text-only channel. Cyc also is trying to store knowledge about a super complicated world in simplistic forms and al... Cyc failed because it lacks natural language. The vast knowledge store of the internet is unintelligible to Cyc. The average person can't use it because they don't speak Cycl and because they have neither the ability nor the patience to translate their implicit thoughts into augmented first order logic. Cyc's approach was understandable when they started in 1984 when they had neither the internet nor the vast computing power that is required to learn natural language from unlabeled examples like children do. Vision and other sensory interpretaion, on the other hand, do not require more info because that... Without natural language, your system
Re: [agi] A Primary Distinction for an AGI
the purpose of text is to convey something. It has to be interpreted. who cares about predicting the next word if you can't interpret a single bit of it. On Tue, Jun 29, 2010 at 3:43 PM, David Jones davidher...@gmail.com wrote: People do not predict the next words of text. We anticipate it, but when something different shows up, we accept it if it is *explanatory*. Using compression like algorithms though will never be able to do this type of explanatory reasoning, which is required to disambiguate text. It is certainly not sufficient for learning language, which is not at all about predicting text. On Tue, Jun 29, 2010 at 3:38 PM, Matt Mahoney matmaho...@yahoo.comwrote: Experiments in text compression show that text alone is sufficient for learning to predict text. I realize that for a machine to pass the Turing test, it needs a visual model of the world. Otherwise it would have a hard time with questions like what word in this ernai1 did I spell wrong? Obviously the easiest way to build a visual model is with vision, but it is not the only way. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Tue, June 29, 2010 3:22:33 PM *Subject:* Re: [agi] A Primary Distinction for an AGI I certainly agree that the techniques and explanation generating algorithms for learning language are hard coded into our brain. But, those techniques alone are not sufficient to learn language in the absence of sensory perception or some other way of getting the data required. Dave On Tue, Jun 29, 2010 at 3:19 PM, Matt Mahoney matmaho...@yahoo.comwrote: David Jones wrote: The knowledge for interpreting language though should not be pre-programmed. I think that human brains are wired differently than other animals to make language learning easier. We have not been successful in training other primates to speak, even though they have all the right anatomy such as vocal chords, tongue, lips, etc. When primates have been taught sign language, they have not successfully mastered forming sentences. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Tue, June 29, 2010 3:00:09 PM *Subject:* Re: [agi] A Primary Distinction for an AGI The point I was trying to make is that an approach that tries to interpret language just using language itself and without sufficient information or the means to realistically acquire that information, *should* fail. On the other hand, an approach that tries to interpret vision with minimal upfront knowledge needs *should* succeed because the knowledge required to automatically learn to interpret images is amenable to preprogramming. In addition, such knowledge must be pre-programmed. The knowledge for interpreting language though should not be pre-programmed. Dave On Tue, Jun 29, 2010 at 2:51 PM, Matt Mahoney matmaho...@yahoo.comwrote: David Jones wrote: I wish people understood this better. For example, animals can be intelligent even though they lack language because they can see. True, but an AGI with language skills is more useful than one without. And yes, I realize that language, vision, motor skills, hearing, and all the other senses and outputs are tied together. Skills in any area make learning the others easier. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Tue, June 29, 2010 1:42:51 PM *Subject:* Re: [agi] A Primary Distinction for an AGI Mike, THIS is the flawed reasoning that causes people to ignore vision as the right way to create AGI. And I've finally come up with a great way to show you how wrong this reasoning is. I'll give you an extremely obvious argument that proves that vision requires much less knowledge to interpret than language does. Let's say that you have never been to egypt, you have never seen some particular movie before. But if you see the movie, an alien landscape, an alien world, a new place or any such new visual experience, you can immediately interpret it in terms of spacial, temporal, compositional and other relationships. Now, go to egypt and listen to them speak. Can you interpret it? Nope. Why?! Because you don't have enough information. The language itself does not contain any information to help you interpret it. We do not learn language simply by listening. We learn based on evidence from how the language is used and how it occurs in our daily lives. Without that experience, you cannot interpret it. But with vision, you do not need extra knowledge to interpret a new situation. You can recognize completely new objects without any training except for simply observing them in their natural state. I wish people understood this better. Dave
Re: [agi] Re: Huge Progress on the Core of AGI
Such an example is no where near sufficient to accept the assertion that program size is the right way to define simplicity of a hypothesis. Here is a counter example. It requires a slightly more complex example because all zeros doesn't leave any room for alternative hypotheses. Here is the sequence: 10, 21, 32 void hypothesis_1() { int ten = 10; int counter = 0; while (1) { print(ten+counter) ten = ten + 10; counter = counter + 1; } } void hypothesis_2() { while (1) print(10 21 32) } Hypothesis 2 is simpler, yet clearly wrong. These examples don't really show anything. Dave On Tue, Jun 29, 2010 at 3:15 PM, Matt Mahoney matmaho...@yahoo.com wrote: David Jones wrote: I really don't think this is the right way to calculate simplicity. I will give you an example, because examples are more convincing than proofs. Suppose you perform a sequence of experiments whose outcome can either be 0 or 1. In the first 10 trials you observe 00. What do you expect to observe in the next trial? Hypothesis 1: the outcome is always 0. Hypothesis 2: the outcome is 0 for the first 10 trials and 1 thereafter. Hypothesis 1 is shorter than 2, so it is more likely to be correct. If I describe the two hypotheses in French or Chinese, then 1 is still shorter than 2. If I describe the two hypotheses in C, then 1 is shorter than 2. void hypothesis_1() { while (1) printf(0); } void hypothesis_2() { int i; for (i=0; i10; ++i) printf(0); while (1) printf(1); } If I translate these programs into Perl or Lisp or x86 assembler, then 1 will still be shorter than 2. I realize there might be smaller equivalent programs. But I think you could find a smaller program equivalent to hypothesis_1 than hypothesis_2. I realize there are other hypotheses than 1 or 2. But I think that the smallest one you can find that outputs eleven bits of which the first ten are zeros will be a program that outputs another zero. I realize that you could rewrite 1 so that it is longer than 2. But it is the shortest version that counts. More specifically consider all programs in which the first 10 outputs are 0. Then weight each program by 2^-length. So the shortest programs dominate. I realize you could make up a language where the shortest encoding of hypothesis 2 is shorter than 1. You could do this for any pair of hypotheses. However, I think if you stick to simple languages (and I realize this is a circular definition), then 1 will usually be shorter than 2. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Tue, June 29, 2010 1:31:01 PM *Subject:* Re: [agi] Re: Huge Progress on the Core of AGI On Tue, Jun 29, 2010 at 11:26 AM, Matt Mahoney matmaho...@yahoo.comwrote: Right. But Occam's Razor is not complete. It says simpler is better, but 1) this only applies when two hypotheses have the same explanatory power and 2) what defines simpler? A hypothesis is a program that outputs the observed data. It explains the data if its output matches what is observed. The simpler hypothesis is the shorter program, measured in bits. I can't be confident that bits is the right way to do it. I suspect bits is an approximation of a more accurate method. I also suspect that you can write a more complex explanation program with the same number of bits. So, there are some flaws with this approach. It is an interesting idea to consider though. The language used to describe the data can be any Turing complete programming language (C, Lisp, etc) or any natural language such as English. It does not matter much which language you use, because for any two languages there is a fixed length procedure, described in either of the languages, independent of the data, that translates descriptions in one language to the other. Hypotheses don't have to be written in actual computer code and probably shouldn't be because hypotheses are not really meant to be run per say. And outputs are not necessarily the right way to put it either. Outputs imply prediction. And as mike has often pointed out, things cannot be precisely predicted. We can, however, determine whether a particular observation fits expectations, rather than equals some prediction. There may be multiple possible outcomes that we expect and which would be consistent with a hypothesis, which is why actual prediction should not be used. For example, the simplest hypothesis for all visual interpretation is that everything in the first image is gone in the second image, and everything in the second image is a new object. Simple. Done. Solved :) right? The hypothesis is not the simplest. The program that outputs the two frames as if independent cannot be smaller than the two frames compressed independently. The program could be made smaller
Re: [agi] A Primary Distinction for an AGI
On Tue, Jun 29, 2010 at 3:33 PM, Mike Tintner tint...@blueyonder.co.ukwrote: You're not getting where I'm coming from at all. I totally agree vision is far prior to language. (We and I've covered your points many times). That's not the point - wh. is that vision is nevertheless still vastly more complex, than you have any idea. whatever you say. That has nothing to do with whether it should be pursued this way or not. For one thing, vision depends on perceptualising/ conceptualising the world - a schematic ontology of the world - image-schematic. It almost certainly has to be done in a certain order, gradually built up. how is that, even remotely, a reason to change the way I do my research? It doesn't even logically follow... No one in our culture has much idea of either what that ontology - a visual ontology - consists of, or how it's built up. Again, how is that an argument for changing my research? It's not. It does not follow again. And for the most basic thing, you still haven't registered that your computer program has ZERO VISION. It's not actually looking at the world at all. It's BLIND - if you take the time to analyse it. A pretty fundamental error/ misconception. Not an argument again. It has nothing to do with whether my approach will or will not provide the valuable knowledge and foundation required to solve the fundamental problems of general vision. Consequently, it also lacks a fundamental dimension of vision, wh. is POINT-OF-VIEW - distance of the visual medium (eg the retina) and viewing subject from the visual object. AGAIN. Not an argument against my approach. It simply doesn't logically follow anything. How is having a point of view in example problems prove that anything learned or developed isn't applicable to general vision? Get thee to a roboticist, make contact with the real world. Get yourself to a psychologist so that they can show you how flawed your reasoning is. Fallacy upon fallacy. You are not in touch with reality. *From:* David Jones davidher...@gmail.com *Sent:* Tuesday, June 29, 2010 6:42 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] A Primary Distinction for an AGI Mike, THIS is the flawed reasoning that causes people to ignore vision as the right way to create AGI. And I've finally come up with a great way to show you how wrong this reasoning is. I'll give you an extremely obvious argument that proves that vision requires much less knowledge to interpret than language does. Let's say that you have never been to egypt, you have never seen some particular movie before. But if you see the movie, an alien landscape, an alien world, a new place or any such new visual experience, you can immediately interpret it in terms of spacial, temporal, compositional and other relationships. Now, go to egypt and listen to them speak. Can you interpret it? Nope. Why?! Because you don't have enough information. The language itself does not contain any information to help you interpret it. We do not learn language simply by listening. We learn based on evidence from how the language is used and how it occurs in our daily lives. Without that experience, you cannot interpret it. But with vision, you do not need extra knowledge to interpret a new situation. You can recognize completely new objects without any training except for simply observing them in their natural state. I wish people understood this better. Dave On Tue, Jun 29, 2010 at 12:51 PM, Mike Tintner tint...@blueyonder.co.ukwrote: Just off the cuff here - isn't the same true for vision? You can't learn vision from vision. Just as all NLP has no connection with the real world, and totally relies on the human programmer's knowledge of that world. Your visual program actually relies totally on your visual vocabulary - not its own. That is the inevitable penalty of processing unreal signals on a computer screen which are not in fact connected to the real world any more than the verbal/letter signals involved in NLP are. What you need to do - what anyone in your situation with anything like your asprations needs to do - is to hook up with a roboticist. Everyone here should be doing that. *From:* David Jones davidher...@gmail.com *Sent:* Tuesday, June 29, 2010 5:27 PM *To:* agi agi@v2.listbox.com *Subject:* Re: [agi] A Primary Distinction for an AGI You can't learn language from language without embedding way more knowledge than is reasonable. Language does not contain the information required for its interpretation. There is no *reason* to interpret the language into any of the infinite possible interpretaions. There is nothing to explain but it requires explanatory reasoning to determine the correct real world interpretation On Jun 29, 2010 10:58 AM, Matt Mahoney matmaho...@yahoo.com wrote: David Jones wrote: Natural language requires more than the words on the page in the real world. Of... Any
Re: [agi] A Primary Distinction for an AGI
Scratch my statement about it being useless :) It's useful, but no where near sufficient for AGI like understanding. On Tue, Jun 29, 2010 at 4:58 PM, David Jones davidher...@gmail.com wrote: notice how you said *context* of the conversation. The context is the real world, and is completely missing. You cannot model human communication using text alone. The responses you would get back would be exactly like eliza. Sure, it might be pleasing to someone that has never seen AI before, but its certainly not answering any questions. This reminds me of the Bing search engine commercials where people ask a question and get responses that include the words they asked about, but in a completely wrong context. Predicting the next word and understanding the question are completely different and cannot be solved the same way. In fact, predicting the next word is altogether useless (at least by itself) in my opinion. Dave On Tue, Jun 29, 2010 at 4:50 PM, Matt Mahoney matmaho...@yahoo.comwrote: Answering questions is the same problem as predicting the answers. If you can compute p(A|Q) where Q is the question (and previous context of the conversation) and A is the answer, then you can also choose an answer A from the same distribution. If p() correctly models human communication, then the response would be indistinguishable from a human in a Turing test. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Tue, June 29, 2010 3:43:53 PM *Subject:* Re: [agi] A Primary Distinction for an AGI the purpose of text is to convey something. It has to be interpreted. who cares about predicting the next word if you can't interpret a single bit of it. On Tue, Jun 29, 2010 at 3:43 PM, David Jones davidher...@gmail.comwrote: People do not predict the next words of text. We anticipate it, but when something different shows up, we accept it if it is *explanatory*. Using compression like algorithms though will never be able to do this type of explanatory reasoning, which is required to disambiguate text. It is certainly not sufficient for learning language, which is not at all about predicting text. On Tue, Jun 29, 2010 at 3:38 PM, Matt Mahoney matmaho...@yahoo.comwrote: Experiments in text compression show that text alone is sufficient for learning to predict text. I realize that for a machine to pass the Turing test, it needs a visual model of the world. Otherwise it would have a hard time with questions like what word in this ernai1 did I spell wrong? Obviously the easiest way to build a visual model is with vision, but it is not the only way. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Tue, June 29, 2010 3:22:33 PM *Subject:* Re: [agi] A Primary Distinction for an AGI I certainly agree that the techniques and explanation generating algorithms for learning language are hard coded into our brain. But, those techniques alone are not sufficient to learn language in the absence of sensory perception or some other way of getting the data required. Dave On Tue, Jun 29, 2010 at 3:19 PM, Matt Mahoney matmaho...@yahoo.comwrote: David Jones wrote: The knowledge for interpreting language though should not be pre-programmed. I think that human brains are wired differently than other animals to make language learning easier. We have not been successful in training other primates to speak, even though they have all the right anatomy such as vocal chords, tongue, lips, etc. When primates have been taught sign language, they have not successfully mastered forming sentences. -- Matt Mahoney, matmaho...@yahoo.com -- *From:* David Jones davidher...@gmail.com *To:* agi agi@v2.listbox.com *Sent:* Tue, June 29, 2010 3:00:09 PM *Subject:* Re: [agi] A Primary Distinction for an AGI The point I was trying to make is that an approach that tries to interpret language just using language itself and without sufficient information or the means to realistically acquire that information, *should* fail. On the other hand, an approach that tries to interpret vision with minimal upfront knowledge needs *should* succeed because the knowledge required to automatically learn to interpret images is amenable to preprogramming. In addition, such knowledge must be pre-programmed. The knowledge for interpreting language though should not be pre-programmed. Dave On Tue, Jun 29, 2010 at 2:51 PM, Matt Mahoney matmaho...@yahoo.comwrote: David Jones wrote: I wish people understood this better. For example, animals can be intelligent even though they lack language because they can see. True, but an AGI with language skills is more useful than one without. And yes, I realize that language, vision, motor skills, hearing
Re: [agi] A Primary Distinction for an AGI
Mike, Alive vs. dead? As I've said before, there is no actual difference. It is not a qualitative difference that makes something alive or dead. It is a quantitative difference. They are both controlled by physics. I don't mean the nice clean physics rules that we approximate things with, I mean the real dynamics of matter. Neither moves any more regularly or irregularly than the other. It is harder to define why something alive moves because the mechanism is normally too complex. If you didn't realize, there are life forms that don't really move, such as viruses. Viruses are controlled by the liquid that contains them. Yet, viruses are arguably alive. Some plants or algae don't really move either. They may just grow in some direction, which is not quite the same as movement. Likewise, your analogy of this to AGI fails. You think there is a difference, but there is none. You may think a fractal is more AGI than a simple, low noise black square, but that is not the case. It is completely besides the point. I can easily add noise to my experiments. I can simulate the noise of light, camera lenses, blurring, etc. But, why should I when, even without noise, there is a clear unsolved AGI challenge. The explanatory reasoning required to solve even zero noise problems is still required for full complexity problems. If you can't solve it for 2 squares on a screen, what makes you think you can solve it for real images? Your grasp of reality regarding AGI is quite poor, in my opinion. Your main claim is that the problems I am working on are not representative or applicable to AGI. But, you fail to see that they really are. The abductive reasoning required to solve these extremely simplified problems is required for every other AGI problem as well. These problems might be solvable using methods that don't apply to AGI. But, that's why it is important to force oneself to solve them in such a way that it IS applicable to AGI. It doesn't mean that you have to choose a problem that is so hard you can't cheat. It's unnecessary to do that unless you can't control your desire to cheat. I can. Developing in this way, such as an implementation of explanatory based reasoning, is very much applicable to AGI. Dave On Mon, Jun 28, 2010 at 11:15 AM, Mike Tintner tint...@blueyonder.co.ukwrote: The recent Core of AGI exchange has led me IMO to a beautiful conclusion - to one of the most basic distinctions a real AGI system must make, and also a simple way of distinguishing between narrow AI and real AGI projects of any kind. Consider - you have a) Dave's square moving across a screen b) my square moving across a screen (it was a sort-of-Pong-player line, but let's make it a square box). How do you distinguish which is animate or inanimate, alive or dead? A very early distinction an infant must make. Remember inanimate objects move (or are moved) too, and in this case you can only see them in motion, - so the self-starting distinction is out. Well, obviously, if Dave's moves *regularly* (like a train or falling stone), it's probably inanimate. If mine moves *irregularly*, - if it stops and starts, or slows and accelerates in irregular, even if only subtly jerky fashion (like one operated by a human Pong player) - it's probably inanimate. That's what distinguishes the movement of life. Inanimate objects normally move *regularly,* in *patterned*/*pattern* ways, and *predictably.* Animate objects normally move *irregularly*, * in *patchy*/*patchwork* ways, and *unbleedingpredictably* . (IOW Newton is wrong - the laws of physics do not apply to living objects as whole objects - that's the fundamental way we know they are living, because they visibly don't obey those laws - they don't normally move regularly like a stone falling to earth, or thrown through the sky. And we're v. impressed when humans like dancers or soldiers do manage by dint of great effort and practice to move with a high though not perfect degree of regularity and smoothness). And now we have such a simple way of distinguishing between narrow AI and real AGI projects. Look at their objects. The really narrow AI-er will always do what Dave did - pick objects that are shaped regularly, move and behave regularly, are patterned, and predictable. Even at as simple a level as plain old squares. And he'll pick closed, definable sets of objects. He'll do this instinctively, because he doesn't know any different - that's his intellectual, logicomathematical world - one of objects that no matter how complex (like fractals) are always regular in shape, movement, patterned, come in definable sets and are predictable. That's why Ben wants to see the world only as structured and patterned even though there's so much obvious mess and craziness everywhere - he's never known any different intellectually. That's why Michael can't bear to even contemplate a world in which things and people behave unpredictably. (And
Re: [agi] A Primary Distinction for an AGI
Yeah. I forgot to mention that robots are not aalive yet could act indistinguishably from what is alive. The concept of alive is likely something that requires inductive type reasoning and generalization to learn. Categorization, similarity analysis, etc could assist in making such distinctions as well. The point is that agi is not defined by any particular problem. It is defined by how you solve problems, even simple ones. Which is why your claim that my problems are not agi is simply wrong. On Jun 28, 2010 12:22 PM, Jim Bromer jimbro...@gmail.com wrote: On Mon, Jun 28, 2010 at 11:15 AM, Mike Tintner tint...@blueyonder.co.uk wrote: Inanimate objects normally move *regularly,* in *patterned*/*pattern* ways, and *predictably This presumption looks similar (in some profound way) to many of the presumptions that were tried in the early days of AI, partly because computers lacked memory and they were very slow. It's unreliable just because we need the AGI program to be able to consider situations when, for example, inanimate objects move in patchy patchwork ways or in unpredictable patterns. Jim Bromer *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
[agi] The true AGI Distinction
In case anyone missed it... Problems are not AGI. Solutions are. And AGI is not the right adjective anyway. The correct word is general. In other words, generally applicable to other problems. I repeat, Mike, you are * wrong*. Did anyone miss that? To recap, it has nothing to do with what problem you solve. It is all about how you solve the problem and your understanding of how the solution is generally applicable to other problems. So, you can kiss it Mike. :D Dave --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] A Primary Distinction for an AGI
I also want to mention that I develop solutions to the toy problems with the real problems in mind. I also fully intend to work my way up to the real thing by incrementally adding complexity and exploring the problem well at each level of complexity. As you do this, the flaws in the design will be clear and I can retrace my steps to create a different solution. The benefit to this strategy is that we fully understand the problems at each level of complexity. When you run into something that is not accounted, you are much more likely to know how to solve it. Despite its difficulties, I prefer my strategy to the alternatives. Dave On Mon, Jun 28, 2010 at 3:56 PM, David Jones davidher...@gmail.com wrote: That does not have to be the case. Yes, you need to know what problems you might have in more complicated domains to avoid developing completely useless theories on toy problems. But, as you develop for full complexity problems, you are confronted with several sub problems. Because you have no previous experience, what tends to happen is you hack together a solution that barely works and simply isn't right or scalable because we don't have a full understanding of the individual sub problems. Having experience with the full problem is important, but forcing yourself to solve every sub problem at once is not a better strategy at all. You may think my strategies has flaws, but I know that and still chose it because the alternative strategies are worse. Dave On Mon, Jun 28, 2010 at 3:41 PM, Russell Wallace russell.wall...@gmail.com wrote: On Mon, Jun 28, 2010 at 4:54 PM, David Jones davidher...@gmail.com wrote: But, that's why it is important to force oneself to solve them in such a way that it IS applicable to AGI. It doesn't mean that you have to choose a problem that is so hard you can't cheat. It's unnecessary to do that unless you can't control your desire to cheat. I can. That would be relevant if it was entirely a problem of willpower and self-discipline, but it isn't. It's also a problem of guidance. A real problem gives you feedback at every step of the way, it keeps blowing your ideas out of the water until you come up with one that will actually work, that you would never have thought of in a vacuum. A toy problem leaves you guessing, and most of your guesses will be wrong in ways you won't know about until you come to try a real problem and realize you have to throw all your work away. Conversely, a toy problem doesn't make your initial job that much easier. It means you have to write less code, sure, but what of it? That was only ever the lesser difficulty. The main reason toy problems are easier is that you can use lower grade methods that could never scale up to real problems -- in other words, precisely that you can 'cheat'. But if you aren't going to cheat, you're sacrificing most of the ease of a toy problem, while also sacrificing the priceless feedback from a real problem -- the worst of both worlds. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] A Primary Distinction for an AGI
Yes I have. But what I found is that real vision is so complex, involving so many problems that must be solved and studied, that any attempt at general vision is beyond my current abilities. It would be like expecting a single person, such as myself, to figure out how to build the h-bomb all by themselves back before it had ever been done. It is the same scenario because it involves many engineering and scientific problems that must all be solved and studied. You see in real vision you have a 3D world, camera optics, lighting issues, noise, blurring, rotation, distance, projection, reflection, shadows, occlusion, etc, etc, etc. It is many magnitudes more difficult than the problems I'm studying. Yet, really consider the two black squares problem. Its hard! It's so simple, yet so hard. I still haven't fully defined how to do it algorithmically... I will get to that in the coming weeks. So, to work on the full problem is practically impossible for me. Seeing as though there isn't a lot of support for AGI research such as this, I am much better served by proving the principle rather than implementing the full solution to the real problem. If I can even prove how vision works on simple black squares, I might be able to get help in my research... without a proof of concept, no one will help. If I can prove it on screenshots, even better. It would be a very significant achievement, if done in a truly general fashion (keeping in mind that truly general is not really possible). A great example of what happens when you work with real images is this... Look at the current solutions. They use features, such as sift. Using sift features, you might be able to say that an object exists with 70% certainty, or something like that. But, it won't be able to tell you what the object looks like, whats behind it. What is it occluding. What's next to it. What color is it. What pixels in the image belong to it. How are those parts attached. Etc. etc. etc. Now do you see why it makes little sense to tackle the full problem? Even the state of the art in computer vision sucks. It is great at certain narrow applications, but no where near where it needs to be for AGI. Dave On Mon, Jun 28, 2010 at 4:00 PM, Russell Wallace russell.wall...@gmail.comwrote: On Mon, Jun 28, 2010 at 8:56 PM, David Jones davidher...@gmail.com wrote: Having experience with the full problem is important, but forcing yourself to solve every sub problem at once is not a better strategy at all. Certainly going back to a toy problem _after_ gaining some experience with the full problem would have a much better chance of being a viable strategy. Have you tried that with what you're doing, i.e. having a go at writing a program to understand real video before going back to black squares and screen shots to improve the fundamentals? --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com