RE: [agi] AGI Preschool: sketch of an evaluation framework for early stage AGI systems aimed at human-level, roughly humanlike AGI
Some might say that if they get conservation of mass and newton's law then they skipped all the useless stuff! OK, but those some probably don't include any preschool teachers or educational theorists. That hypothesis is completely at odds with my own intuition from having raised 3 kids and spent probably hundreds of hours helping out in daycare centers, preschools, kindergartens, etc. Sorry, that was just kind of a joke. Probably nobody actually has the opinion I was lampooning though I do see similar things said sometimes, as if inferring minimum-description-length root level reductionisms is a realistic approach to learning to deal with the world. It might even be true, but the humor was supposed to be to juxtapose that idea with the AGI preschool. --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
RE: [agi] AGI Preschool: sketch of an evaluation framework for early stage AGI systems aimed at human-level, roughly humanlike AGI
Ben: Right. My intuition is that we don't need to simulate the dynamics of fluids, powders and the like in our virtual world to make it adequate for teaching AGIs humanlike, human-level AGI. But this could be wrong.I suppose it depends on what kids actually learn when making cakes, skipping rocks, and making a mess with play-dough. Some might say that if they get conservation of mass and newton's law then they skipped all the useless stuff! I think I agree with the plausibility of something you have said many times: that there may be many paths to AGI that are not similar at all to human development -- abstract paths to modelling the universe, teasing meaning from sheer statistics of the chinese/chinese dictionary of the raw html internet, who knows what. But in the case where we are trying to roughly follow stages of human development with goals of producing human-like linguistic and reasoning capabilities, I very much fear that any significant simplification of the universe will provide an insufficient basis for the large sensory concept set underlying language and analogical reasoning (both gross and fine). Literally, I think you're throwing the baby out with the bathwater. But, as you say, this could be wrong. It's really the only critique I have of the AGI preschool idea, which I do like because we can all relate to it very easily. At any rate, if it turns out to be a valid criticism the symptom will be that an insufficiently rich set of concepts will develop to support the range of capabilities needed and at that point the simulations can be adjusted to be more complete and realistic and provide more human sensory modalities. I guess it will be disappointing if building an adequate virtual world turns out to be as difficult and expensive as building high quality robots -- but at least it's easier to clean up after cake-baking. --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
RE: [agi] AGI Preschool: sketch of an evaluation framework for early stage AGI systems aimed at human-level, roughly humanlike AGI
Oh, and because I am interested in the potential of high-fidelity physical simulation as a basis for AI research, I did spend some time recently looking into options. Unfortunately the results, from my perspective, were disappointing. The common open-source physics libraries like ODE, Newton, and so on, have marginal feature sets and frankly cannot scale very well performance-wise. Once I even did a little application whose purpose was to see whether a human being could learn to control an ankle joint to compensate for an impulse event and stabilize a simple body model (that is, to make it not fall over) by applying torques to the ankle. I was curious to see (through introspection) how humans learn to act as process controllers. http://happyrobots.com/anklegame.zip for anybody bored enough to care. It wasn't a very good test of the question so I didn't really get a satisfactory answer. I did discover, though, that a game built around more appealing cases of the player learning to control physics-inspired processes could be quite absorbing. Beyond that, the most promising avenue seems to be physics libraries tied to graphics hardware being worked on by the hardware companies to help sell their stream processors. The best example is Nvidia, who bought PhysX and ported it to their latest cards, giving a huge performance boost. Intel has bought Havok and I can only imagine that they are planning on using that as the interface to some Larrabee-based physics engine. I'm sure that ATI is working on something similar for their newer (very impressive) stream processing cards. At this stage, though, despite some interesting features and leaping performance, it is still not possible to do things like get realistic sensor maps for a simulated soft hand/arm, and complex object modifications like bending and breaking are barely dreamed of in those frameworks. Complex multi-body interactions (like realistic behavior when dropping or otherwise playing with a ring of keys or realistic baby toys) have a long ways to go. Basically, I fear those of us who are interested in this are just waiting to ride the game development coattails and it will be a few years at least until performance that even begins to interest me will be available. Just my opinions on the situation. --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
RE: [agi] AGI Preschool: sketch of an evaluation framework for early stage AGI systems aimed at human-level, roughly humanlike AGI
Hi Ben. OTOH, if one wants to go the virtual-robotics direction (as is my intuition), then it is possible to bypass many of the lower-level perception/actuation issues and focus on preschool-level learning, reasoning and conceptual creation. And yet, in your paper (which I enjoyed), you emphasize the importance of not providing a simplistic environment (with the screwdriver example). Without facing the low-level sensory world (either through robotics or through very advanced simulations feeding senses essentially equivalent to those of humans), I wonder if a targeted human-like AGI will be able to acquire the necessary concepts that children absorb and use as much o f the metaphorical basis for their thought -- slippery, soft, hot, hard, rough, sharp, and on and on. I assume you have some sort of middle ground in mind... what's your thinking about how much you can cheat in this way (beyond that of what is conveniently doable I mean)? Thanks! --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
RE: [agi] The Future of AGI
Although a lot of AI-type research focuses on natural language interfaces between computer systems and their human users, computers have the ability to create visual images (which people can't do in real-time beyond gestures and facial expressions). Building computer systems that generate pictures or videos as their way of communicating with us could be a very lucrative addition to computer applications that include cognitive models of their users (instead of focusing solely on generating natural language), because most of us do process visual information so well. This is really narrow AI I suppose, though it's kind of on the borderline. It does seem like one of the ways to commercialize incremental progress toward AGI. Derek Zahn supermodelling.net --- 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=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] Hunting for a Brainy Computer
Pei Wang: --- I have problem with each of these assumptions and beliefs, though I don't think anyone can convince someone who just get a big grant that they are moving in a wrong direction. ;-) With his other posts about the Singularity Summit and his invention of the word Synaptronics, Modha certainly seems to be a kindred spirit to many on this list. I think what he's trying to do with this project (to the extent I understand it) seems like a reasonably promising approach (not really to AGI as such, but experimenting with soft computing substrates is kind of a cool enterprise to me). Let a thousand flowers bloom. However, when he says things on his blog like In my opinion, there are three reasons why the time is now ripe to begin to draw inspiration from structure, dynamics, function, and behavior of the brain for developing novel computing architectures and cognitive systems. -- I despair again. Dr. Wang, if you want to get some funding maybe you should start promoting NARS as a theory of the brain :) --- 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=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] A paper that actually does solve the problem of consciousness
Richard, As a general rule, I find discussions about consciousness, qualia, and so forth to be unhelpful, frustrating, and unnecessary. However, I enjoyed this paper a great deal. Thanks for writing it. Because of my inclinations on these matters, I am not an expert on the history of thought on the topic, or its current status among philosophers, but I find your account to be credible and reasonably clear. I'm not particularly repulsed by the idea that ... our most immediate, subjective experiance of the world is, in some sense, an artifact produced by the operation of the brain so searching for a more satisfying conclusion is not really high up on my priority list. Still, I don't see anything immediately objectionable in your analysis. I am not certain about the distinguishing power of your falsifiable predictions, but only because I would need to give that considerably more thought. I look forward to being in the audience when you present the paper at AGI-09. Derek Zahn agiblog.net --- 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=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] A paper that actually does solve the problem of consciousness
Oh, one other thing I forgot to mention. To reach my cheerful conclusion about your paper, I have to be willing to accept your model of cognition. I'm pretty easy on that premise-granting, by which I mean that I'm normally willing to go along with architectural suggestions to see where they lead. But I will be curious to see whether others are also willing to go along with you on your generic cognitive system model. --- 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=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: AW: AW: [agi] If your AGI can't learn to play chess it is no AGI
Matthias Heger: If chess is so easy because it is completely described, complete information about state available, fully deterministic etc. then the more important it is that your AGI can learn such an easy task before you try something more difficult. Chess is not easy. Becoming good at chess is something that most humans never accomplish and none accomplish without years of training in background material. The question is whether chess is representative of the domains we want AGIs to master. I think a case could be made either way. I don't want to be discouraging -- any concrete demonstration of AGI ideas is of great interest, even in formal toy domains. --- 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=117534816-b15a34 Powered by Listbox: http://www.listbox.com
RE: [agi] META: A possible re-focusing of this list
As somebody who considers consciousness, qualia, and so on to be poorly-defined anthropomorphic mind-traps, I am not interested in any such discussions. Other people are, and I have no problem ignoring them, like I ignore a number of individual cranks and critics who post things of similarly low interest.I think a forum divided into topic areas would be better than this mailing list for many different reasons, but if you don't want to move to that setup and if you want to police the posts more actively (this list, according to agiri.org, is already supposed to be about technical aspects of particular AGI approaches), it won't bother me. I do like to see different perspectives on issues of common interest if they are of high quality. That is subjective, though. For example, I consider Matt Mahoney and Richard Loosemore to contribute very interesting material, even if I do not agree with their conclusions. Others may consider Mike Tintner and Steve Richfield to have useful things to say, when I do not. Date: Wed, 15 Oct 2008 15:18:14 -0400From: [EMAIL PROTECTED]: [EMAIL PROTECTED]: Re: [agi] META: A possible re-focusing of this list By the way, I'm avoiding responding to this thread till a little time has passed and a larger number of lurkers have had time to pipe up if they wish to...ben --- 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=117534816-b15a34 Powered by Listbox: http://www.listbox.com
RE: [agi] META: A possible re-focusing of this list
I bet if you tried very hard to move the group to the forum (for example, by only posting there yourself and periodically urging people to use it), people could be moved there. Right now, nobody posts there because nobody else posts there; if one wants one's stuff to be read, one sends it to the high traffic location unless there's a reason not to. Date: Wed, 15 Oct 2008 16:00:45 -0400From: [EMAIL PROTECTED]: [EMAIL PROTECTED]: Re: [agi] META: A possible re-focusing of this list There is already a forum site on agiri.org . Nobody uses it So, just setting up a forum site is not the answer...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/?member_id=8660244id_secret=117534816-b15a34 Powered by Listbox: http://www.listbox.com
RE: [agi] META: A possible re-focusing of this list
How about this: Those who *do* think it's worthwhile to move to the forum: Instead of posting email responses to the mailing list, post them to the forum and then post a link to the response to the email list, thus encouraging threads to continue in the more advanced venue. I shall do this myself from now on. I have not participated much on this list lately due to my current work schedule but will make an effort to do so. If used, I do think the forum could help solve some of these META issues. --- 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=117534816-b15a34 Powered by Listbox: http://www.listbox.com
RE: [agi] META: A possible re-focusing of this list
Oh, also: When I try to register a form account, it says:Sorry, an error occurred. If you are unsure on how to use a feature, or don't know why you got this error message, try looking through the help files for more information. The error returned was: To register, please send your request to [EMAIL PROTECTED] Please include your desired username.A random password will be sent back to you. --- A forum that won't let people register isn't likely to catch on. --- 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=117534816-b15a34 Powered by Listbox: http://www.listbox.com
RE: [agi] Advocacy Is no Excuse for Exaggeration
I am reminded of this: http://www.serve.com/bonzai/monty/classics/MissAnneElk Date: Tue, 14 Oct 2008 17:14:39 -0400From: [EMAIL PROTECTED]: [EMAIL PROTECTED]: Re: [agi] Advocacy Is no Excuse for Exaggeration OK, but you have not yet explained what your theory of consciousness is, nor what the physical mechanism nor role for consciousness that you propose is ... you've just alluded obscurely to these things. So it's hard to react except with raised eyebrows and skepticism!!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/?member_id=8660244id_secret=117534816-b15a34 Powered by Listbox: http://www.listbox.com
RE: [agi] A NewMetaphor for Intelligence - the Computer/Organiser
It has been explained many times to Tintner that even though computer hardware works with a particular set of primitive operations running in sequence, a hardwired set of primitive logical operations operating in sequence is NOT the theory of intelligence that any AGI researchers are proposing (to my knowledge). A computer is just a system for holding a theory of intelligence which does not look like those primitives (at least not since the view that intelligence consists of simple interpretations of atomic tokens representing physical objects in small numbers of relationships with other such tokens was given up decades ago as insufficient). As an example, the representational mechansms in Novamente and the dynamics of the mind agents that operate on them are probably better thought of as churning masses of probability relationships with varying and often non-specific semantic interpretations than Tintner's narrow view of what a computer is -- although I do not yet understand Novamente in detail. He has to ignore all such efforts, though, because if he paid attention he would have to stop saying that NONE of us understand ANYTHING about how REAL intelligence is actually based on line drawings, or keyboards, or other childish notions. Though he's in my killfile I do see his posts when others take the bait. So Mike, please try to finally understand this: AGI researchers do not think of intelligence as what you think of as a computer program -- some rigid sequence of logical operations programmed by a designer to mimic intelligent behavior. We know it is deeper than that. This has been clear to just about everybody for many many years. By engaging the field at such a level you do nothing worthwhile. Date: Sat, 6 Sep 2008 15:38:59 +0100 From: [EMAIL PROTECTED] To: agi@v2.listbox.com Subject: Re: [agi] A NewMetaphor for Intelligence - the Computer/Organiser 2008/9/6 Mike Tintner [EMAIL PROTECTED]: Will, Yes, humans are manifestly a RADICALLY different machine paradigm- if you care to stand back and look at the big picture. Employ a machine of any kind and in general, you know what you're getting - some glitches (esp. with complex programs) etc sure - but basically, in general, it will do its job. What exactly is a desktop computers job? --- 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=111637683-c8fa51 Powered by Listbox: http://www.listbox.com
RE: [agi] The Necessity of Embodiment
By embodied I think people usually mean a dense sensory connection (with a feedback loop) to the physical world. The feedback could be as simple as aiming a camera. However, it seems to me that an AI program connected to YouTube could maybe have a dense enough link to the real world to charge up a grounded sufficiently-complete and human-compatible set of concepts. A large quantity of video of other intelligences interacting with the world could maybe substitute for direct interaction. --- 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=111637683-c8fa51 Powered by Listbox: http://www.listbox.com
RE: [agi] OpenCog Prime wikibook and roadmap posted (moderately detailed design for an OpenCog-based thinking machine)
Ben, Thanks for the large amount of work that must have gone into the production of the wikibook. Along with the upcoming PLN book (now scheduled for Sept 26 according to Amazon) and re-reading The Hidden Pattern, there should be enough material for a diligent student to grok your approach. I think it will take some considerable time for anybody to absorb it all, so don't be too discouraged if there isn't a lot of visible banter about issues you think are important; we all come at the Big Questions of AGI from our own peculiar perspectives. Even those of us who want to believe may have difficulty finding sufficient common ground in viewpoints to really understand your ideas in depth, at least for a while. If there's one thing I'd like to see more of sometime soon, it would be more detail on the early stages of your vision of a roadmap, to help focus both analysis and development. Great stuff! --- 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=108809214-a0d121 Powered by Listbox: http://www.listbox.com
RE: [agi] Simple example of the complex systems problem, for those in a hurry
Thanks again Richard for continuing to make your view on this topic clear to those who are curious. As somebody who has tried in good faith and with limited but nonzero success to understand your argument, I have some comments. They are just observations offered with no sarcasm or insult intended. 1) The presentations would be a LOT clearer if you did not always start with Suppose that... and then make up a hypothetical situation. As a reader I don't care about the hypothetical situation, and it is frustrating to be forced into trying to figure out if it is somehow a metaphor for what I *am* interested in, or what exactly the reason behind it is. In this case, if you are actually talking about a theory of how evolution produced a significant chunk of human cognition (a society of CBs), then just say so and lead us to the conclusions about the actual world. If you are not theorizing that the evolution/CBs thing is how human minds work, then I do not see the benefit of walking down the path. Note that the basic CB idea you user here strikes me as a good one; it resonates with things like Minsky's Society of Mind, as well as the intent behind things like Hall's Sigmas and Goertzel's subgraphs. 2) Similarly, when you say if we were able to look inside a CB system and see what the CBs are doing [Note: we can do this, to a limited extent: it is called introspection], we would notice many aspects of CB behavior ... It would be a lot better if you left out the if and the would. Say when we look inside this CB system... and we do notice any aspects... if that is what you mean. If again this is some sort of strange hypothetical universe as a reader I am not very interested in speculations about it. 3) When you say But now, here is a little problem that we have to deal with. It turns out that the CB system built by evolution was functioning *because* of all that chaotic, organized mayhem, *not* in spite of it. Assuming that you are actually talking about human minds instead of a hypothetical universe, this is a very strong statement. It is a theory about human intelligence that needs some support. It is not necessarily a theory about intelligence-in-general; linking it to intelligence in general would be another theory requring support. You may or may not think that intelligence in general is a coherent concept; given your recent statements that there can be no formal definition of intelligence, it's hard to say whether intelligence that is not isomorphic to human intelligence can exist in your view. 4) Regarding: Evolution explored the space of possible intelligent mechanisms. In the course of doing so, it discovered a class of systems that work, but it may well be that the ONLY systems in the whole universe that can function as well as a human intelligence involve a small percentage of weirdness that just balances out to make the system work. There may be no cleaned-up versions that work. The natural response is: sure, this may well be, but it just as easily may well not be. This is addressed in your concluding points, which say that it is not definite, but is very likely. As a reader, I do not see a reason to suppose that this is true. You offer only the circumstantial evidence that AI has failed for 50 years, but there are many other possible reasons for this: - maybe it's just hard. many aspects of the universe took more than 50 years to understand, many are still not understood. i personally think that if this is true we are unlikely to be just a few years from the solution, but it does seem like a reasonable viewpoint. - maybe logic just stinks as a tool for modeling the world. it seemed natural but looking at the things and processes in the human universe logically seems like a pretty poor idea to me. maybe probabilistic logic of one sort or another will help. but the point here is that it might not be a complex systems issue, it might just be a knowledge representation and reasoning issue. perhaps generated or evolved program fragments will fare better; perhaps things that look like neural clusters will work, perhaps we haven't discovered a good way to model the universe yet. - maybe we haven't ripped the kitchen sink out of the wall yet... maybe intelligence will turn out to be a conglomeration of 1543 different representation schemes and reasoning tricks, but we've only put a fraction together so far and therefore only covered a small section of what intelligence needs to do. 5) Of course, the argument would be strengthened by a somewhat detailed suggestion of how AI research *should* proceed; you give some arguments for why certain (unspecified) approaches *might* not work, but nothing beyond the barest hint of what to do about it, which doesn't motivate anybody to give much more than a shrug to your comments. I wonder what it is that you expect people to do in response to
RE: [agi] Simple example of the complex systems problem, for those in a hurry
Oh, one last point: I find your thoughts in this message quite interesting personally because I think that puzzling out exactly what concept builders need to do, and how they might be built to do it, is the most interesting thing in the whole world. I am resistant to the idea that it is impossible because all efforts to do so must be destined to result in insufficient results. I admit to stubbornness on this point, and it will take strong deprogramming to stop me from taking an interest in recipes for the philosophers' stone. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
RE: [agi] Simple example of the complex systems problem, for those in a hurry
Sorry for three messages in short succession. Regarding concept builders, I have been writing in my bumbling way about this (and will continue to muse on fundamental issues) in my little blog: http://agiblog.net --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
RE: [agi] WHAT SORT OF HARDWARE $33K AND $850K BUYS TODAY FOR USE IN AGI
I agree that the hardware advances are inspirational, and it seems possible that just having huge hardware around could change the way people think and encourage new ideas. But what I'm really looking forward to is somebody producing a very impressive general intelligence result that was just really annoying because it took 10 days of computing instead of an hour. Seems to me that all the known AGI researchers are in theory, design, or system building phases; I don't think any of them are CPU-bound at present -- and no fair pointing to Goedel Machines or AIXI either, which will ALWAYS be resource-starved :) --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
RE: [agi] Approximations of Knowledge
Richard, If I can make a guess at where Jim is coming from: Clearly, intelligent systems CAN be produced. Assuming we can define intelligent system well enough to recognize it, we can generate systems at random until one is found. That is impractical, however. So, we can look at the problem as one of search optimization. Evolution produced intelligent systems through a biased search, for example, so it is at least possible to improve search over completely random generate and test. What other ways can be used to speed up search? Jim is suggesting some methods that he believes may help. If I understand what you've said about your approach, you have some very different methods than what he is proposing to focus the search. I do not understand exactly what Jim is proposing; presumably he is aiming to use his SAT solver to guide the search toward areas that contain partial solutions or promising partial models of some sort. It seems to me very difficult to define the goal formally, very difficult to develop a meta system in which a sufficiently broad class of candidate systems can be expressed, and very difficult to describe the splices or reductions or partial models in such a way to smooth the fitness landscape and thus speed up search. So I don't know how practical such a plan is. But (again assuming I understand Jim's approach) it avoids your complex system arguments because it is not making any effort to predict global behavior from the low-level system components, it's just searching through possibilities. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
[agi] Roadrunner PetaVision
Brain modeling certainly does seem to be in the news lately. Checking out nextbigfuture.com, I was reading about that petaflop computer Roadrunner and articles about it say that they are or will soon be emulating the entire visual cortex -- a billion neurons. I'm sure I'm not the only one who thinks that knowing what the cortex does and roughly how it does it could be quite inspiring for AGI, so I was surprised at this news. Does anybody have links to more information (besides the short recent mainstream news story)? Are they just being enthusiastic about their big computer or do they have a sophisticated theory? --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
RE: [agi] Definition of AGI - comparison with animals
Dr. Matthias Heger: Which animal has the smallest level of intelligence which still would be sufficient for a robot to be an AGI-robot? You ask for opinions, we got lots of those! I believe most people on this list would consider that humans are the only animals with significant-enough amounts of general intelligence to warrant the label. For example, using Goertzel's definition for intelligence: complex goals in complex environments -- the goals of non-human animals do not seem complex in the same way that building an airplane is complex... although it is interesting whether, say, becoming the dominant member of my group for other primates is really a simple goal. In any case, the generality of goals that can be undertaken by nonhuman animals seems very limited. Human cognition evolved from less capable forms. Do our higher cognitive functions depend on the lower neural machinery so much that it does not make sense to talk about abstract thought without the base functions shared by other animals? Figuring that out is a good reason to study animal-level cognition even though it is not the ultimate goal. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
RE: [agi] IBM, Los Alamos scientists claim fastest computer
TeslasTwo things I think are interesting about these trends in high-performance commodity hardware: 1) The flops/bit ratio (processing power vs memory) is skyrocketing. The move to parallel architectures makes the number of high-level operations per transistor go up, but bits of memory per transistor in large memory circuits doesn't go up. The old bit per op/s or byte per op/s rules of thumb get really broken on things like Tesla (0.03 bit/flops). Of course we don't know the ratio needed for de novo AGI or brain modeling, but the assumptions about processing vs memory certainly seem to be changing. 2) Much more than previously, effective utilization of processor operations requires incredibly high locality (processing cores only have immediate access to very small memories). This is also referred to as arithmetic intensity. This of course is because parallelism causes operations per second to expand much faster than methods for increasing memory bandwidth to large banks. Perhaps future 3D layering techniques will help with this problem, but for now AGI paradigms hoping to cache in (yuk yuk) on these hyperincreases in FLOPS need to be geared to high arithmetic intensity. Interestingly (to me), these two things both imply to me that we get to increase the complexity of neuron and synapse models beyond the muladd/synapse + simple activation function model with essentially no degradation in performance since the bandwidth of propagating values between neurons is the bottleneck much more than local processing inside the neuron model. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
RE: [agi] Pearls Before Swine...
Gary Miller writes: We're thinking Don't feed the Trolls! Yeah, typical trollish behavior -- upon failing to stir the pot with one approcah, start adding blanket insults. I put Steve Richfield in my killfile a week ago or so, but I went back to the archive to read the message in question. The reason it got no response is that it is incoherent. Seriously, I couldn't even understand the point of it. Something about dreams and brains being wired completely different and some thumbnail calculations which are not included but apparently conclude that AGI will need the entire population of the earth for software maintenance... um, that's just weird rambling crackpottery. It is so far away from any sort of AGI nuts and bolts that it cannot even be parsed. There are people who do not believe they are crackpots (but are certainly perceived that way) who then transform into trolls spouting vague blanket insults and whining about being ignored. That type of unsupported fringe wackiness is tolerated because, frankly, the whole field is fringe to most people. When it turns into vague attacks, blanket condemnation, and insults (a la Tintner and now Richfield) it simply isn't worth reading any more. For others in danger of spiraling down the same drain, I recommend: * Be cordial. Note: condescending is not cordial. * Be specific and concise. Stick to one point. * Do not refer to decades-old universally ignored papers about character recognition as if they are AI-shaping revolutions. * Do not drop names from some hazy good old days * Attempt to limit rambling off-topic insights into marginally related material * If you are going to criticize instead of putting forward positive ideas (why you'd bother criticizing this field is beyond me, but if you must): criticize specific things, not the herd or all of you researchers or the field of AGI... as Ben pointed out earlier, no two people in this area agree on much of anything and they cannot be lumped together. Criticizing specific things means actually reading and attempting to understand the published works of AGI researchers -- the test for whether you belong here is whether you are willing and able to actually do that. Mr. Richfield may find a more receptive audience here: http://www.kurzweilai.net/mindx/frame.html --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
RE: [agi] Ideological Interactions Need to be Studied
Speaking of neurons and simplicity, I think it's interesting that some of the how much cpu power needed to replicate brain function arguments use the basic ANN model, assuming a MULADD per synapse, updating at say 100 times per second (giving a total computing power of about 10^16 OPS). But the people actually trying brain replication (I am thinking of the Blue Brain cortical column project) simulate 30 million synapses on a 22 TF computer, four orders of magnitude higher (~1M op/sec/synapse vs 100). Markram is certainly a rabid optimist in terms of how far this work will go and how fast, so I don't think he's just throwing cycles away for no reason. Assuming their neuron model turns out to be adequate, I wonder how far it can be squeezed down. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
RE: [agi] More Info Please
Mark Waser: Does anybody have any interest in and/or willingness to program in a different environment? I haven't decided to what extent I'll participate in OpenCog myself yet. For me, it depends more on whether the capabilities of the system seem worth exploring, which in turn depends as much on the underlying philosophy as the codebase. I'm thinking of OpenCog right now as a concrete way to understand the ideas of Bencompany. Frankly I find OpenCog a rather odd open source project given its open-ended nature -- no target end users, no clear applications, no (apparent) dedicated driving personality declaring here is exactly what we need to accomplish, who's with me?. I don't mean that Ben isn't dedicated, but I don't envision him herding this particularly ornery flock and browbeating people into actually finishing and debugging code. Still, it's a very cool effort wherever it leads. The language used doesn't particularly matter to me, so I'm willing to work in a different environment. I don't have a Linux machine at the moment so a requirement to work in Linux is a small but significant barrier to entry for me. Screwing around with operating systems is just about my least favorite thing to do.It's hard for me even to make my own guesses about the best way to go because the overall architecture isn't very clear to me yet. I guess that the central data structure -- the AtomTable, contains a persistent cached bunch nodes and links that come in various types and have numbers attached to them. But it's not clear to me whether the types are supposed to be part of the cognitive theory or not -- are OpenCog developers supposed to invent new node types or just work with those provided? If they can be created, does that mean changing the AtomTable implementation? Is the meaning of the numbers on links predetermined or can they be overloaded? If they can be overloaded, how do the Mind Agents cope with the ensuing chaos? If they can't be overloaded, how can the system be extended to include new ideas? If for example the AtomTable is sufficiently compartmentalized so that it won't need changing, it would seem that porting it to another language would be a lower priority than providing an environment where developing Mind Agents (which I am assuming is the really interesting stuff) could occur in whatever language individual developers feel most productive in. Or maybe the amount of work required to do even this is larger than the actual interest in using the code warrants. The more interesting issues to me are things like how the Atoms (and any other representational structures I don't know about yet) get their semantics and how adding new code changes those semantics... what is the representational flexibility and power of the knowledge representation scheme when applied to some non-toy cases... I'm sure the documentation will make things a lot clearer. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
RE: [agi] Re: Merging - or: Multiplicity
Steve Richfield: It is sure nice that this is a VIRTUAL forum, for if we were all in one room together, my posting above would probably get me thrashed by the true AGI believers here. Does anyone here want to throw a virtual stone? Sure. *plonk* --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
RE: [agi] Pattern extrapolation as a method requiring limited intelligence
John Rose writes: So I feel that much of our brain mass is there due to the natural richness of nature, and there may be quite a bit of overkill compared to what would be needed in software AGI. Are we satisfied building AGIs that cannot cope with the actual world because it is too rich? Personally I think that without the natural richness of nature, our intelligence would never develop. We climb those levels of richness like a rock face. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
RE: [agi] Pattern extrapolation as a method requiring limited intelligence
Vladimir Nesov: I think sterile texture of artificial environments hides the richness of their structure from our intuition, since we already have it imprinted by experience with the real world. Anything less than capable of dealing with the real world won't understand cleaned up environments also. +1. John Rose: Which actual world, a natural or manmade? Both, at least up to the present day. In my opinion (though I know from your previous post that you don't agree), I don't see a huge difference in the environmental complexity of the land on which New York City sits now vs 1000 years ago. I did not grow up, nor do I live, in a mostly featureless box. I do agree with your more general point that SOME of the brain's functionality does not have to be duplicated in silicon to achieve AGI. Whether it is a significant fraction, and whether it would need to be replaced with some other functionality, seems like a hard question to me. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
[agi] AI in virtual worlds -- popular press
For those who might not have seen it yet, seems this concept is becoming rather popular: http://www.msnbc.msn.com/id/24668099/ --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
RE: Symbol Grounding [WAS Re: [agi] AGI-08 videos]
Richard Loosemore writes: some very useful text about the symbol grounding problem. Thank you Richard. For once I don't feel like a complete idiot. I am familiar with these Harnad papers and find them quite clear. Beyond that I understand your further explanation and even agree personally that it is a critical issue. It is especially interesting in view of the push to embody AGI projects in virtual worlds, about which I am spending some time thinking anyway. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] AGI-08 videos
Richard Loosemore: So, for example, if I were organizing a conference on AGI I would want people to address such questions as: I find your list of questions to be quite fascinating, and I'd love to participate in an active list or conference devoted to these Foundations of Cognitive Computing type of issues. However, it doesn't particularly bother me that people are building systems without explicit answers to these things, because I find the systems themselves, and the ideas about AI that they embody, to be very cool on their own terms. I am not competing with any of them for money or fame, I am hopeful that lessons will be learned no matter how right or wrong their approaches end up, I think we're decades away from AGI systems that are intelligent enough to have a real impact on our society (a more useful phrase that than human level IMO) so I'm not mad that wasted effort is delaying a cure for hunger and disease. People do not accept critiques that cast their entire professional output as worthless and their most basic premises as fatally flawed... if a point can be made in an understandable way from the assumed world view of somebody else, I do think it's worth making, but it's somewhat rare to be able to do so on material with which I have only a casual familiarity. The deep questions that interest you (and me and, to an extent i believe, everybody on this list) are troublesome because they are so hard to talk about. Consider your complex systems argument. There appears to be some basic point-of-view differences that make communication on these topics difficult. It's not all pigheadedness or ill will, I don't think. Or (picking one of your questions at random): - What assumptions do we have to buy into if we go with bayesian nets as a choice of reasoning/representation formalism? And how would we go about finding out if those assumptions are valid enough to make it safe to use bayes nets? I'm not sure how to even begin a conversation about such a question. First we have to decide what a reasoning/representation formalism has to do, and I'm afraid everybody has a different set of premises on points like that. Those debates would be highly worthwhile, but I doubt many people will bother with them. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
[agi] AGI-08 videos
I noticed yesterday that most of the videos of talks and panels from AGI-08 have been uploaded (http://www.agi-08.org/schedule.php). Big thanks to the organizers for that! I have some difficulty getting into some of the papers but the 10-ish minute overview talks are by and large quite good, and the panel discussions are particularly interesting. I feel much better now about not going to the conference! Hopefully the rest of the talks will be posted, I can't wait to watch them. Some personal reactions to particular things: * Finally, I think the field has moved beyond the need for so many papers on six secrets of AI, five reasons AI has failed, and so on. Even the obligatory What is AI? talk was largely redundant (although Dr. Wang's point -- that we will all have different definitions and we should take that into account when studying the work of others -- needs saying). This is good news. Perhaps next year's conference won't need so many overview talks. * Somehow I had this vague notion that SOAR had basically dried up and blown away in the 1990s, but John Laird's description of current work in SOAR was terrific and quite exciting. I'll be following their progress closely. * There are now quite a number of architectures with AGI-type ambitions that have significant implementation behind them (Novamente, SOAR, LIDA, NARS, OSCAR, BICA, Texai, and others). The most interesting parts of the panels for me was when the people involved in building those architectures discussed what they have in common and differences in approach for similar problems. As Ben Goertzel (and Sam Adams and others) point out, these architectures share quite a lot at the level of their boxes and arrows overview slides, which provides some context for interesting detailed discussion. If such discussion occurred on this list that would be really cool; but perhaps a workshop at AGI-09 where the architects of these actually-existing systems discussed their similarities and differences and current limitations would be worthwhile. I'd sure pay to see it! * It was quite interesting to see that simulation/visualization as an important operating principle / reasoning mechanism is becoming so popular. Ideally, I suppose, such modal mechanisms would do double duty in perception and simulation... accomplishing that and interfacing it cleanly with other modalities or general-purpose knowledge representation is really fascinating and I have a feeling we'll be seeing more along those lines. I wonder if Novamente will go sort of solipsistic and absorb AGISim into itself as a modal reasoning module. * Along those lines, there seems to be a growing (certainly not universal) consensus among complete-system builders that virtual embodiment is a best approach for providing broad knowledge support (grounding) without messing around with robots. Somebody could write an excellent paper about the potential pitfalls of such an approach (detail, fidelity, deep causality issues behind appearance, function, and inter-object + inter-feature relationships, and so on). If nobody else is working in detail on publishing such an analysis perhaps I will study those issues for some months and try to write something for AGI-09 about it. * Stephen Reed is one of the most clear and deliberate speakers I've seen in this field. It's really interesting how seeing a person talk about their research makes it seem more real and interesting than just reading about it. * I wish Josh's Variac paper wasn't just a poster... but I suppose something has to get left out. Hopefully next year there will be more concrete implementation/experimentation progress to report in a talk. * Limiting people to 10-12 minutes makes it basically impossible to present the contents of a paper, so the talks turn into project overviews. Actually I found that to be a GOOD thing, and hope it continues that way (as long as we don't get the same overview talks year after year...) * Some presenters were very effective and some were not. I encourage everybody to rehearse their talks to make sure that the amount of material presented is appropriate to the time frame, and to make sure it is presented smoothly. Thanks to the organizers and all the participants. Fantastic stuff. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] AGI-08 videos
One other observation I forgot to mention: Several people brought up the desirability of some kind of benchmark problem area to help compare the methods and effectiveness of various approaches. For a bunch of reasons I think it will be difficult to define such things in a way that researchers will pay attention to, but if it could be done (either as simply a commonly-understood point of reference for discussions or as a grand challenge or anything in between) it would be very neat, and in my opinion beneficial to the field as a whole. I have a suggestion for such a task: figuring out how to operate the buttons-and-light system that determines whose turn it is to talk during panel discussions. It may be too ambitious though, as clearly it requires superhuman intelligence (har har har). --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] AGI-08 videos
Bob Mottram writes: I havn't watched all of the AGI-08 videos, but of those that I have seen the 15 minute format left me non the wiser. With limited time I would have preferred longer talks with more depth but perhaps fewer in number, especially on the more mathematical topics. Another suggestion. If there is an AGI-09 perhaps part of the conference could be in Second Life, allowing for longer discussion if needed. If you wanted to get really fancy you could set up a projector and incorporate speakers/questioners from within SL as part of the live conference. Interesting ideas. I'm not sure the participants would have preferred 10 in-depth math lectures to the format actually chosen, but I agree that explaining papers with sufficient time would be better than what was possible to actually do. I guess the only way to achieve that is break into parallel topic sessions... though it does seem unfortunate for a field with general in its name to split up into special interests. The second life thing seems like a good idea, though I don't see how it helps the time issue. Perhaps a pre-first-life or post-first-life meeting (a couple weeks before after the conference) in second life for detailed lectures about the papers could work, on a relaxed schedule. Most people (especially scientists) have some access to a video camera these days; I wonder if a longer talk could be videotaped at home by the participants and then put up on google video for later viewing at leisure, in addition to the on-site filming. I don't know if it's standard practice at conferences these days to videotape the sessions and make them available like this, but it's really wonderful. Bob, if you wrote a paper for the conference about state of the art vision or robotic systems and how they relate to AGI research, that would be very cool I think! --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] AGI-08 videos
Richard Loosemore writes: Prompted by your enthusiastic write-up, I just wasted one and a half hours scanning through all of the AGI-08 papers that I downloaded previously. I have 28 of them; they did not include anything from Stephen Reed, nor any NARS paper, so I guess my collection must be incomplete, but even so I saw absolutely nothing that makes me believe that a field called Artificial General Intelligence even exists yet. Sorry for wasting your time Richard! At least you got to keep your collection of doomed projects up to date! There was no NARS paper but Pei Wang did have a paper called What Do You Mean By AI? that I'm sure you'd find horrifying. Stephen Reed's talk was called Natural Language Approach of the Texai Project, in which he described in some detail what he has been implementing. Surprisingly, I don't see a paper with that title in the paper list. My post was only about the videos of the talks/panels and the impression they gave regarding the conference as a whole. I'm not sure I'd bother with the videos though if I were you... I have a feeling you would find them unsatisfying. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] AGI-08 videos
Richard Loosemore: I read Pei's paper and there was nothing horrifying about it (please spare the sarcasm). No sarcasm intended. If I had just come to the conclusion that 28 papers in a row were a waste of time, I'd be horrified at the prospect of a 29th that would also not give me what I was looking for. I was merely trying for a light tone while expressing my belief that you were unlikely to change your conclusion about the field of Artificial General Intelligence by spending further time with that paper or the videos. I shall, however, spare you such glib language in future correspondence. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] AGI-08 videos
Richard Loosemore: My god, Mark: I had to listen to people having a general discussion of grounding (the supposed them of that workshop) without a single person showing the slightest sign that they had more than an amateur's perspective on what that concept actually means. I was not at that workshop and am no expert on that topic, though I have seen the word used in several different ways. Could you point at a book or article that does explain the concept or at least use it heavily in a correct way? I would like to improve my understanding of the meaning of the grounding concept. Note: sometimes written words do not convey intensions very well -- I am not being sarcastic, I am asking for information to help improve the quality of discussion that you have found lacking in the past. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] help me,please for books for agi and mind in pdf
Bruno Frandemiche asked for online AGI-related text. If you're adventurous, I'd recommend the Workshop proceedings from 2006: http://www.agiri.org/wiki/Workshop_Proceedings and the conference proceedings from AGI-08: http://www.agi-08.org/papers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] An interesting project on embodied AGI
Thanks, what an interesting project. Purely on the mechanical side, it shows how far away we are from truly flexible house-friendly robust mobile robotic devices. I'm a big fan of the robotic approach myself. I think it is quite likely that dealing with the messy flood of dirty data coming from the real world, and sorting it all into forms where it can be learned from effectively, will be very helpful in rooting out the nature of concepts and the meaning of symbols as applied to the actual world (rather than some non-representative abstraction). On the other hand, if we want to build philosophers and scientists, it would be kind of nice if we didn't have to build them expensive bodies! Do you have to be able to think like a cat to think like Descartes? I'm glad people are looking for the answer from both sides of the issue. A quote: Whatever the robot learns individually and socially should help to develop its language skills, which in turn, should help iCub interact better with its environment and pick up more knowledge to learn more. Knowledge of grammar and vocabulary is also likely to emerge naturally through this process. I wonder if they have a theory of mind that would make this actually possible, or if they are just optimistic. Sounds sort of magical to me so far. If I ever actually start experimenting with AGI for my own amusement, I'll start by building a robot. Of course, I've been saying that for years, and I am a big fan of Bob Mottram's work in this area. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] How general can be and should be AGI?
I assume you are referring to Mike Tintner. As I described a while ago, I *plonk*ed him myself a long time ago, most mail programs have the ability to do that. and it's a good idea to figure out how to do it with your own email program. He does have the ability to point at other thinkers and their papers, such as Lakoff and Barsalou, who have extremely interesting things to say... but his own contributions (beyond citing) to any converation are infuriating., I think it's about time to give up on Mike until he learns to behave again. And you shouldn't use sarcasm -- he just doesn't get it. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] Why Symbolic Representation P.S.
The little Barsalou I have read so far has been quite interesting, and I think there are a lot of good points there, even if it is a rather extreme position. The issue of how concepts (which is likely a nice suitcase word lumping a lot of discrete or at least overlapping cognitive functions into one blob) originate, either from sensory data or from other concepts, is the most interesting thing in the world to me and trying to think about it from all angles, although quite time consuming, is a fun way to spend evenings. I think that the following response to a particularly-important Barsalou article would resonate with the viewpoints of many people on this list: http://www.vub.ac.be/CLEA/liane/Reviews/Barsalou.htm --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] WHAT ARE THE MISSING CONCEPTUAL PIECES IN AGI? --- recent input and responses
J Andrew Rogers writes: Most arguments and disagreements over complexity are fundamentally about the strict definition of the term, or the complete absence thereof. The arguments tend to evaporate if everyone is forced to unambiguously define such terms, but where is the fun in that. I agree with this to a point at least. My attempt to rephrase Richard's argument falters because I have not yet understood his use of the term 'complexity'. I'd prefer a rigorous definition but will settle for a better general understanding of what he means. Despite his several attempts to describe his meaning I have not been able yet to successfully grasp exactly what counts as complex and what does not, and for things inbetween, how to judge the degree of complexity. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] WHAT ARE THE MISSING CONCEPTUAL PIECES IN AGI? --- recent input and responses
Richard: I get tripped up on your definition of complexity: A system contains a certain amount of complexity in it if it has some regularities in its overall behavior that are governed by mechanisms that are so tangled that, for all practical purposes, we must assume that we will never be able to find a closed-form explanation of how the global arises from the local.on figuring out what counts as a regularity in overall behavior. Consider a craps table. The trajectories of the dice would seem to have global regularities (for which craps players and normal people have words and phrases, like bouncing off the back, flying off the table, or whatever). Our ability to create concepts around this activity would seem to imply the existence of global regularities (finding them is what we do when we make concepts). Yet the behavior of those regularities is not just physical law but the specific configuration of the felt, the chips, the wind, and so forth, and all that data makes a closed-form explanation impractical. Yet, I don't get the sense that this is what you mean by a complex system. If it is, your contention that they are rare is certainly not correct, since many such examples can easily be found. This aspect of complexity iillustrates the butterfly effect often used in discussions of complexity. I'm not trying to be difficult; it's crucial for me to understand what you mean (versus my interpretation of what others have meant or my own internal definitions) if I am to follow your argument. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: Thoughts on the Zahn take on Complex Systems [WAS Re: [agi] WHAT ARE THE MISSING ...]
Mark Waser: Huh? Why doesn't engineering discipline address building complex devices? Perhaps I'm wrong about that. Can you give me some examples where engineering has produced complex devices (in the sense of complex that Richard means)? --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: Thoughts on the Zahn take on Complex Systems [WAS Re: [agi] WHAT ARE THE MISSING ...]
Me: Can you give me some examples where engineering has produced complex devices (in the sense of complex that Richard means)? Mark: Computers. Anything that involves aerodynamics. Richard, is this correct? Are human-engineered airplanes complex in the sense you mean? --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: Thoughts on the Zahn take on Complex Systems [WAS Re: [agi] WHAT ARE THE MISSING ...]
Mark Waser: I don't know what is going to be more complex than a variable-geometry-wing aircraft like a F-14 Tomcat. Literally nothing can predict it's aerodynamic behavior. The avionics are purely reactive because it's future behavior cannot be predicted to any certainty even at computer speeds -- yet it's behavior envelope is small enough to be safe, provided you do have computer speeds (though no human can fly it unaided). I agree that this is a very sensible way to think about being complex and it is certainly similar to the way I think about it myself. My embryonic understanding of Richard's argument suggests to me that he means something else, though. If not, traditional engineering methods are often pretty good at taming complexity as long as they take the range of possible system states into account (which is what you have been saying all along). Since I'm trying (with limited success) to understand his point of view, I might suggest that (from the point of view of his argument), the global regularities of the aircraft (its flight characteristics) DO have a sufficiently-efficacious small theory in terms of the components (the aircraft body, including the moveable bits). In fact, it is exactly that small theory which is embedded in the control program. Since the global regularities (straight-line flight, turns, and so on) are sufficiently predictable from the local interactions of the control surfaces with the air, the aircraft is not complex *in the sense that Richard is talking about*. Now I suppose I've pissed everybody off, but I'm really just trying to understand Richard's definitions so I can follow his argument. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: Thoughts on the Zahn take on Complex Systems [WAS Re: [agi] WHAT ARE THE MISSING ...]
Richard Loosemore: it makes no sense to ask is system X complex?. You can only ask how much complexity, and what role it plays in the system. Yes, I apologize for my sloppy language. When I say is system X complex? what I mean is whether the RL-complexity of the system is important in describing the behaviors of interest under the operating conditions being discussed, in particular whether the global behaviors have an effective small theory expressed in terms of local components and their interactions -- because my current understanding of what you mean by complexity means the extent to which no such small theory is available. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] For robotics folks: Seeking thoughts about integration of OpenSim and Player
Ben Goertzel writes: it might be valuable to have an integration of Player/Stage/Gazebo with OpenSim I think this type of project is a good start toward addressing one of the major critiques of the virtual world approach -- the temptation to (unintentionally) cheat -- those canned animations can fool the untrained observer into thinking there is a lot more going on than they would think from the actual agi output (MOVE AVATAR 3M FORWARD, e.g.). Similarly for input... if input is obtained from simulated senses instead of providing direct access to the object model of the simulator, the resulting internal representations in the AI are much more convincing somehow. I don't know much about OpenSim or Player/Stage/Gazebo but it would seem that they overlap quite a bit, in that they both include simulation as their core function. I guess what you really want to do is figure out a middle layer between the standard player interface and the specific virtual world of interest (OpenSim). My guess would be that you could re-use a bunch of code as part of that task but it will still be rather complex. It would be an interesting GSoC project I'd think. It doesn't address the other major criticisms of virtual worlds -- such as the fact that in current ones the objects are more iconic than realistic... by which I mean that the appearance and limited behavior of a mug of beer is purely for the purpose of being identifiable by humans with a lot of experience of real mugs of beer. Its role in the virtual world is suggestive rather than functional and in a lot of cases I have no idea how an AI could make any useful sense out of it. I imagine that criticism will be dealt with over a lot of time as the worlds become more realistic, though the computing requirements for something approaching the richness of real-world sensation seem daunting, as does the sheer volume of object data needed to realistically model something as simple as a baby's bedroom. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] WHAT ARE THE MISSING CONCEPTUAL PIECES IN AGI? --- recent input and responses
One more bit of ranting on this topic, to try to clarify the sort of thing I'm trying to understand. Some dude is telling my AGI program: There's a piece called a 'knight'. It moves by going two squares in one direction and then one in a perpendicular direction. And here's something neat: Except for one other obscure case I'll tell you about later, it's the only piece that moves by jumping through the air instead of moving a square at a time on its journey. When I try to think about how an intelligence works, I wonder about specific cases like these (and thanks to William Pearson for inventing this one) -- the genesis of the knight concept from this specific purely verbal exchange. How could this work? What is it about the specific word sequences and/or the conversational context that creates this new thing -- the Knight? It would have to be a hugely complicated language processing system... so where did that language processing system come from? Did somebody hardcode a model of language and conversation and explicitly insert generate concept here actions? That sounds like a big job. If it was learned (much better), how was it learned? What is the internal representation of the language processing model that leads to this particular concept formation, and how was it generated? If I can see something specific like that in a system (say Novamente) I can start to really understand the theory of mind it expresses. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] WHAT ARE THE MISSING CONCEPTUAL PIECES IN AGI? --- recent input and responses
Stephen Reed writes: Hey Texai, let's program [Texai] I don't know how to program, can you teach me by yourself? Sure, first thing is that a program consists of statements that each does something [Texai] I assume by program you mean a sequence of instructions that a computer can interpret and execute, and that by statement you mean a line of code written as part of a computer program Right. One type of instruction is the assignment statement. [Texai] OK, how is it interpreted? It has two parts, one part identifies a variable that receives a copy of, or a reference to, an evaluated expression described by the other part of the instruction [Texai] I assume by variable you mean a symbol (like x or y) that is used in mathematical or logical expressions to represent a variable quantity. What's an evaluated expression? And under what circumstances does the situation in which the variable receives a copy of the evaluated expression occur, as contrasted with the situation in which the variable receives a reference to the evaluated expression? Wow, if that turns out to be an actual transcript sent back through a time machine (I mean, if it works like you think), that's amazingly impressive. Every part of it, from knowing to ask you to teach it to do something, to connecting 'program' used as a verb to 'program' used as a noun, to knowing all about sequences of instructions, what computers are and how they work, what a line of code even means, and so on. I assume these things were taught to it through previous teaching sessions, and I'm really eager to see that in action. Of particular interest to me here is the conceptual leap from equality in a mathematical expression (which I guess the system already knows about) to the very different idea of assignment in a normal programming language. The origin of a variable as a named thing that can hold a value was an interesting concept to communicate to undergraduate business majors back in the day when I taught introductory programming... you could just see them get it after trying analogies with mailboxes and diagrams of computer memory and whatnot. It had never occurred to some of them to put a number in a box for later use before but I clearly remember the instant of concept formation occurring in their fresh young minds :) Now the aha moment behind learning the concept of recursion is even more interesting... --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] WHAT ARE THE MISSING CONCEPTUAL PIECES IN AGI? --- recent input and responses
Vladimir Nesov writes: Generating concepts out of thin air is no big deal, if only a resource-hungry process. You can create a dozen for each episode, for example. If I am not certain of the appropriate mechanism and circumstances for generating one concept, it doesn't help to suggest that a dozen get generated instead... now I have twelve times as many things to explain. If you are suggesting that concept formation is a (perhaps stochastic) generate-and-test procedure, that seems like an okay idea but the issues are then redescribed as: what is the generation procedure, what causes it to be invoked, what the test procedure is, and so on. These questions cannot be answered outside the context of a particular system; they are just the things I'd like to understand exactly how they would happen in Novamente or Texai or whatever, with all handwaving removed. To get back to the original question of this thread, these are some of the many missing conceptual pieces TO ME because I cannot see the specific nuts and bolts solution for any proposed system. It may in fact be that for any non-toy example the mechanisms and data are going to be too complicated for such analysis... that is, my brain is too puny and ineffective to understand (in a clear and relatively complete way) the inner workings of a general intelligence. In that case, all I can do is hope for proof by performance. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] WHAT ARE THE MISSING CONCEPTUAL PIECES IN AGI? --- recent input and responses
Richard Loosemore: I do not laugh at your misunderstanding, I laugh at the general complacency; the attitude that a problem denied is a problem solved. I laugh at the tragicomedic waste of effort. I'm not sure I have ever seen anybody successfully rephrase your complexity argument back at you; since nobody understands what you mean it's not surprising that people are complacent about it. I was going to wait for some more blog posts to have a go at rephrasing it myself but my (probably wrong effort) would go like this: 1. Many things we want to build have desired properties that are described at a different level than the things we build them out of. Flying is emergent in this sense from rivets and sheet metal, for example. Thinking is emergent from neurons, for another example. 2. Some such things are complex in that the emergent properties cannot be predicted from the lower-level details. 3. Flying as above is not complex in this way. In fact, all of engineering is the study of how to build things that are increasingly complicated but NOT complex. We do not want airplanes to have complex behavior and the engineering methodology is expressly for squeezing complexity out. 4. Thinking must be complex. [my understanding of why this must be true is lacking. Something like: otherwise we'd be able to predict the behavior of an AGI which would make it useless?] 5. Therefore we have no methods for building thinking machines, since engineering discipline does not address how to build complex devices. Building them as if they are not complex will result in poor behavior; squeezing out the complexity will squeeze out the thinking, and leaving it in makes traditional engineering impossible. Not quite right I suppose, but I'll keep working at it. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] WHAT ARE THE MISSING CONCEPTUAL PIECES IN AGI?
Josh writes: You see, I happen to think that there *is* a consistent, general, overall theory of the function of feedback throughout the architecture. And I think that once it's understood and widely applied, a lot of the architectures (repeat: a *lot* of the architectures) we have floating around here will suddenly start working a lot better. Want to share this theory? :) Oh, by the way, of the ones I read so far, I thought your Variac paper was the most interesting one from AGI-08. I'm particularly interested to hear more about sigmas and your thoughts on transparent, composable, and robust programming languages. I used to think about some slightly related topics and thought more in terms of evolvability and plasticity (and did not consider opaqueness at all) but I think your approach to thinking about things is quite exciting. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: Thoughts on the Zahn take on Complex Systems [WAS Re: [agi] WHAT ARE THE MISSING ...]
Richard Loosemore: I'll try to tidy this up and put it on the blog tomorrow. I'd like to pursue the discussion and will do so in that venue after your post. I do think it is a very interesting issue. Truthfully I'm more interested in your specific program for how to succeed than this argument about why everybody else will fail, but I understand that they are linked. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] WHAT ARE THE MISSING CONCEPTUAL PIECES IN AGI?
William Pearson writes: Consider an AI learning chess, it is told in plain english that... I think the points you are striving for (assuming I understand what you mean) are very important and interesting. Even the first simplest steps toward this clear and (seemingly) simple task baffle me. How does the concept of 'knight' poof into existence during the conversation? How does a system learn how to learn to play a game in the first place? I like this task as a tool for considering how a potential AGI approach is truly general -- by asking over and over again how and why could that happen for any imagining of how each sentence could be processed. Now, Edward, I hope you are right about Novamente but I don't quite follow the reasoning behind your confidence. I'm imagining that in a previous life you'd pointed me toward a drawing of a DaVinci flying machine, excitedly projecting 3-8 years until we'd be flying around. Now DaVinci's a bright guy (smarter than me) and it's a nice concept, and I can't prove it won't work -- I'd have to invent a pretty effective aerodynamic science to do so. I still might not be convinced. Absence of disproof is not necessarily strong evidence. I'm looking forward to getting more info about Novamente soon and hopefully understand the nuts and bolts of how it could do tasks like the ones William wrote about. I have some concerns about things like whether propagating truth values around is really a very effective modeling substrate for the world of objects and ideas we live in -- but since I don't understand Novamente well enough, there's lititle I can say pro or con beyond those vague intuitions (and the last thing I'd want to do is bug Ben with questions like how would Novamente do X? How about Y? He has plenty of real work to do.) --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] associative processing
Steve Richfield writes: Hmm, I haven't seen a reference to those core publications. Is there a semi-official list? This list is maintained by the Artificial General Intelligence Research Instutute. See www.agiri.org . On that site there are several semi-official lists -- under Publications and Instead of an AGI Textbook. Certainly there is very little agreement (on anything!) amongst the idiosyncratic group of people who post on this list and I did not intend to dissuade you from presenting your ideas (which I have found interesting so far, in proportion to the degree they address AGI topics); I was just explaining why people here are unlikely to find Dr. Eliza to be particularly interesting. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] associative processing
Note that the Instead of an AGI Textbook section is hardly fleshed out at all at this point, but it does link to a more-complete similar effort to be found here: http://nars.wang.googlepages.com/wang.AGI-Curriculum.html --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] associative processing
Steve Richfield, writing about J Storrs Hall: You sound like the sort that once the things is sort of roughed out, likes to polish it up and make it as good as possible. I don't believe your characterization is accurate. You could start with this well-done book to check that opinion: http://www.amazon.com/Beyond-AI-Creating-Conscience-Machine/dp/1591025117 Because you are new to the discussion here you probably don't quite get the topic of this mailing list (AGI); the system sort-of described in your papers does not address any of the issues of that topic (as defined in its core publications and conferences) so don't be too surprised if people here are not particularly excited about it. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] Logical Satisfiability...Get used to it.
Jim Bromer writes: With God's help, I may have discovered a path toward a method to achieve a polynomial time solution to Logical Satisfiability If you want somebody to talk about the solution, you're more likely to get helpful feedback elsewhere as it is not a topic that most of us on this list deal with or know a lot about. Besides that, publish your result and it will be used if it is helpful. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
RE: [agi] Intelligence: a pattern discovery algorithm of scalable complexity.
[EMAIL PROTECTED] writes: But it should be quite clear that such methods could eventually be very handy for AGI. I agree with your post 100%, this type of approach is the most interesting AGI-related stuff to me. An audiovisual perception layer generates semantic interpretation on the (sub)symbolic level. How could a symbolic engine ever reason about the real world without access to such information? Even more interesting: How could a symbolic engine ever reason about the real world *with* access to such information? :) --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
RE: [agi] Intelligence: a pattern discovery algorithm of scalable complexity.
Stephen Reed writes: How could a symbolic engine ever reason about the real world *with* access to such information? I hope my work eventually demonstrates a solution to your satisfaction. Me too! In the meantime there is evidence from robotics, specifically driverless cars, that real world sensor input can be sufficiently combined and abstracted for use by symbolic route planners. True enough, that is one answer: by hand-crafting the symbols and the mechanics for instantiating them from subsymbolic structures. We of course hope for better than this but perhaps generalizing these working systems is a practical approach. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
[agi] Symbols
Related obliquely to the discussion about pattern discovery algorithms What is a symbol? I am not sure that I am using the words in this post in exactly the same way they are normally used by cognitive scientists; to the extent that causes confusion, I'm sorry. I'd rather use words in their strict conventional sense but I do not fully understand what that is. These thoughts are fuzzier than I'd like; if I was better at de-fuzzifying them I might be a pro instead of an amateur! Proposition: a symbol is a token with both denotative and model-theoretic semantics. The denotative semantacs are what makes a symbol refer to something or be about something. The model-theoretic semantics allow symbol processing operations to occur (such as reasoning). I believe this is a somewhat more restrictive use of the word symbol than is necessarily implied by Newell and Simon in the Physical Symbol System Hypothesis, but my aim is engineering rather than philosophy. I'm actually somewhat skeptical that human beings use symbols in this sense for much of our cognition. We appear to be a million times better at it than any other animal, and that is the special thing that makes us so great, but we still aren't very good at it. However, most of the things we want to build AGI *for* require us to greatly expand the symbol processing capabilities of mere humans. I think we're mostly interested in building artificial scientists and engineers rather than artificial musicians. Since computer programs, engineering drawings, and physics theories are explicitly symbolic constructs, we're more interested in effectively creating symbols than in the totality of the murky subsymbolic world supporting it. To what extent can we separate them? I wish I knew. In this view, subsymbolic simply refers to tokens that lack some of the features of symbols. For example, a representation of a pixel from a camera has clear denotational semantics but it is not elaborated as well as a better symbol would be (the light coming from direction A at time B is not as useful as the light reflecting off of Fred's pinky fingernail). Similarly, and more importantly, subsymbolic products of sensory systems lack useful model-theoretic semantics. The origin of symbols problem involves how those semantics arise -- and to me it's the most interesting piece of the AGI puzzle. Is anybody else interested in this kind of question, or am I simply inventing issues that are not meaningful and useful? --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
RE: [agi] Intelligence: a pattern discovery algorithm of scalable complexity.
Mark Waser writes: True enough, that is one answer: by hand-crafting the symbols and the mechanics for instantiating them from subsymbolic structures. We of course hope for better than this but perhaps generalizing these working systems is a practical approach. Um. That is what is known as the grounding problem. I'm sure that Richard Loosemore would be more than happy to send references explaining why this is not productive. It's not the grounding problem. The symbols crashing around inthese robotic systems are very well grounded. The problem is that these systems are narrow, not that they manipulateungrounded symbols. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
[agi] Novamente study
Ben, It seems to me that Novamente is widely considered the most promising and advanced AGI effort around (at least of the ones one can get any detailed technical information about), so I've been planning to put some significant effort into understanding it with a view toward deciding whether I think you're on the right track or not (with as little hand-waving, faith, or bigotry as possible in my conclusion). To do that properly, I am waiting for your book on Probabilistic Logic Networks to be published. Amazon says July 2008... is that date correct? Thanks! --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
RE: [agi] Novamente study
Ben Goertzel writes: The PLN book should be out by that date ... I'm currently putting in some final edits to the manuscript... Also, in April and May I'll be working on a lot of documentation regarding plans for OpenCog. Thanks, I look forward to both of these. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
RE: [agi] Complexity in AGI design
Dennis Gorelik writes: Derek, I quoted this Richard's article in my blog: http://www.dennisgorelik.com/ai/2007/12/reducing-agi-complexity-copy-only-high.html Cool. Now I'll quote your blogged response: So, if low level brain design is incredibly complex - how do we copy it? The answer is: we don't copy low level brain design. Low level design is critical for AGI. Instead we observe high level brain patterns and try to implement them on top of our own, more understandable, low level design. I'm not sure for myself what I think of this complexity argument, so I don't have anything to say about your answer except to wish you luck (if Richard is right, you'll need a lot of it; if many paths lead up the hill then you might not need much at all). I am curious what you mean by high level brain patterns though. Could you give an example? - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=73825873-cc7440
RE: [agi] Solution to Grounding problem
Richard Loosemore writes: This becomes a problem because when we say of another person that they meant something by their use of a particular word (say cat), what we actually mean is that that person had a huge amount of cognitive machinery connected to that word cat (reaching all the way down to the sensory perception mechanisms that allow the person to recognise an instance of a cat, and motor output mechanisms that let them interact with a cat). What Stephen Harnad said in his original paper was Hang on a second: if the AI system does not have all that other machinery inside it when it uses a word like cat, surely it does not really mean the same thing by cat as a person would? [...] Thanks, Richard. That post was a terrific bit of writing. On a related note, I think those that are uneasy with the idea of grounding symbols in experience with a virtual world wonder whether the (current) thin and skewed sensory experiece of cats or any other concept-friendly regularities in such worlds are sufficiently similar to provide enough of the same meaning for communication with humans using the resulting concepts. For that matter, one wonders even when concepts are grounded in the real world whether the resulting concepts and their meanings can be similar enough for communication if the concept formation machinery is not quite similar to our own sometimes even individual human conceptualizations are barely similar enough to allow conversation. Very interesting stuff. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=73797095-e37936
RE: [agi] None of you seem to be able ...
Richard Loosemore writes: Okay, let me try this. Imagine that we got a bunch of computers [...] Thanks for taking the time to write that out. I think it's the most understandable version of your argument that you have written yet. Put it on the web somewhere and link to it whenever the issue comes up again in the future. If you are right, you may have to resort to told you so when other projects fail to produce the desired emergent intelligence. No matter what you do, system builders can and do and will say that either their system is probably not heavily impacted by the issue, or that the issue itself is overstated for AGI development, and I doubt that most will be convinced otherwise. By making such a clear exposition, at least the issue is out there for people to think about. I have no position myself on whether Novamente (for example) is likely to be slain by its own complexity, but it is interesting to ponder. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=73249587-454993
RE: [agi] What best evidence for fast AI?
Hi Robin. In part it depends on what you mean by fast. 1. Fast - less than 10 years. I do not believe there are any strong arguments for general-purpose AI being developed in this timeframe. The argument here is not that it is likely, but rather that it is *possible*. Some AI researchers, such as Marvin Minsky, believe that we already have the necessary hardware commonly available, if we only knew what software to write for it. If, as seems likely, there is a large economic incentive for the development of this software, it seems reasonable to grant the possibility that it will be developed. Following that line of reasoning, a computation of probability * impact yields a large number for even small probabilities since the impact of a technological singularity could be very large. So planning for the possibility seems prudent. 2. Fast - less than 50 years. For this timeframe, just dust off Moravec's old computer speed chart. On such a chart I think we're supposed to be at something like mouse level right now -- and in fact we have seen supercomputers beginning to take a shot at simulating mouse-brain-like structures. It does not feel so wrong to think that the robot cars succeeding in the DARPA challenges are maybe up to mouse-level capabilities. It is certainly possible that once computers surpass the raw processing power of the human brain by 10, 100, 1000 times, we will just be too stupid to keep up with their capabilities for some reason, but it seems like a more reasonable bet to me that the economic pressures to make somewhat good use of available computing resources will win out. AI is often called a perpetual failure, but from this view that is not true at all; AI has been a spectacular success. It's very impressive that the early researchers were able to get computers with nematode-level nervous systems to show any interesting cognitive behavior at all. At worst, AI is keeping up with the available machine capabilities admirably. Still, putting aside the brain simulation route, we do have to build models of mind that actually work. As Pei Wang just pointed out, we are beginning to see models such as Ben Goertzel's Novamente that at least seem like they might have a shot at sufficiency. That is not proof, but it is an indication that we may not be overmatched by this challenge, once the machinery becomes available. If something like Moore's law continues (I suppose it's a cognitive bias to assume it will continue and a different bias to assume it won't), who wants to bet that computers 10,000, 100,000, or 1,000,000 times as powerful as our brains will go to waste? Add as many zeros as you want... they cost five years each. - Having written that, I confess it is not completely convincing. There are a lot of assumptions involved. I don't think there *is* an objectively convincing argument. That's why I never try to convince anybody... I can play in the intersection between engineering and wishful thinking if I want, simply because it amuses me more than watching football. Hopefully some folks with more earnest beliefs will have better arguments for you. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=63831279-12920a
RE: [agi] What best evidence for fast AI?
Bryan Bishop: Looks like they were just simulating eight million neurons with up to 6.3k synapses each. How's that necessarily a mouse simulation, anyway? It isn't. Nobody said it was necessarily a mouse simulation. I said it was a simulation of a mouse-brain-like structure. Unfortunately, not enough is yet known about specific connectivity so the best that can be done is play with structures of similar scale in anticipation of further advances. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=63847412-4e7cf3
RE: [agi] How valuable is Solmononoff Induction for real world AGI?
Edward, For some reason, this list has become one of the most hostile and poisonous discussion forums around. I admire your determined effort to hold substantive conversations here, and hope you continue. Many of us have simply given up. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=63054363-e5048c
RE: [agi] Connecting Compatible Mindsets
A large number of individuals on this list are architecting an AGI solution (or part of one) in their spare time. I think that most of those efforts do not have meaningful answers to many of the questions, but rather intend to address AGI questions from a particular perspective. Would such people be encouraged to fill this out, even though they might only answer a couple of the numbered points? Probably most people like that are not serious contenders in the sense of having a complete detailed plan for achieving a full AGI. Rather they think a particular aspect or approach is not being given enough attention and hope to explore part of it to see if their ideas merit further development. I could see wanting to include or exclude such amateur efforts, depending on the goals of this database. Perhaps a separate section would be a good idea for such people to provide a brief unstructured summary of their interests and ideas. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=62431770-893502
RE: [agi] Poll
1. What is the single biggest technical gap between current AI and AGI? I think hardware is a limitation because it biases our thinking to focus on simplistic models of intelligence. However, even if we had more computational power at our disposal we do not yet know what to do with it, and so the biggest gap is conceptual rather than technical. In particular, I become more and more skeptical that the effort to produce concise theories of things like knowledge representation are likely to succeed. Frames, is-a relations, logical inference on atomic tokens, and so on, are efforts to make intelligent behavior comprehensible in concisely describable ways, but they seem to only be crude approximations to the reality of intelligent behavior, which seem less and less likely to have formulations that are comfortably within our human ability to reason about effectively. As one example, consider the study in cognitive science of the theory of categories -- from the necessary and sufficient conditions classical view to the more modern competing views of prototypes vs exemplars. All of these are nice simple descriptions but as so often happens it seems that the effort to boil down the phenomena to nice simple ideas we can work with in our tiny brains actually boils off most of the important stuff. The challenge is for us to come up with ways to think about or at least work with (and somehow reproduce or invent!) mechanisms that appear not to be reduceable to convenient theories. I expect that our ways of thinking about these things will evolve as the systems we build operate on more and more data. As Novamente's atom table grows from thousands to millions and eventually billions of rows; as cortex simulations become more and more detailed and studyable; as we start to grapple with semantic nets containing many millions of nodes -- our understanding of the dynamics of such systems will increase. Eventually we will become comfortable with and become more able to build systems whose desired behaviors cannot even be specified in a simple or rigorous way. Or, perhaps, theoretical breakthroughs will occur making it possible to describe intelligence and its associated phenomena in simple scientific language. Because neither of these things can be done at present, we can barely even talk to each other about things like goals, semantics, grounding, intelligence, and so forth... the process of taking these unknown and perhaps inherently complex things and compressing them into simple language symbols throws out too much information to even effectively communicate what little we do understand. Either way, it will take decades if we're lucky. Moving from mouse-level hardware to monkey-level hardware in the next couple decades will be helpful, just like our views on machine intelligence have expanded beyond those of our forebears looking at the first digital computers and wondering about how they might be made to think. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=55002867-d97b38
RE: Self-improvement is not a special case (was Re: [agi] Religion-free technical content)
Tim Freeman writes: Let's take Novamente as an example. ... It cannot improve itself until the following things happen: 1) It acquires the knowledge and skills to become a competent programmer, a task that takes a human many years of directed training and practical experience. 2) It is given access to its own implementation and permission to alter it. 3) It understands its own implementation well enough to make a helpful change. ... I agree that resource #1, competent programming, is essential for any interesting takeoff scenario. I don't think the other two matter, though. Ok, this alternative scenario -- where Novamente secretly reinvents the theoretical foundations needed for AGI development, designs its successor from those first principles, and somehow hijacks an equivalent or superior supercomputer to receive the de novo design and surreptitiously trains it to superhuman capacity -- should also be protected against. It's a fairly ridiculous scenario, but for completeness should be mentioned. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=53003223-9d4579
RE: Self-improvement is not a special case (was Re: [agi] Religion-free technical content)
Tim Freeman: No value is added by introducing considerations about self-reference into conversations about the consequences of AI engineering. Junior geeks do find it impressive, though. The point of that conversation was to illustrate that if people are worried about Seed AI exploding, then one option is to not build Seed AI (since that is only one approach to developing AGI, and in fact I do not know of any actual project that includes it at present). Quoting Yudkowsky: The task is not to build an AI with some astronomical level of intelligence; the task is building an AI which is capable of improving itself, of understanding and rewriting its own source code. Perhaps only junior geeks like him find the concept relevant. You seem to think that self-reference buys you nothing at all since it is a simple matter for the first AGI projects to reinvent their own equivalent from scratch, but I'm not sure that's true. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=53018769-30e88d
RE: Self-improvement is not a special case (was Re: [agi] Religion-free technical content)
Linas Vepstas: Let's take Novamente as an example. ... It cannot improve itself until the following things happen:1) It acquires the knowledge and skills to become a competent programmer, a task that takes a human many years of directed training and practical experience. Wrong. This was hashed to death in previous emails; and then again probably several more times before I joined the list. Anyone care to assemble a position paper on self improvement that reviews the situation? I'm slightly irritated by the recurring speculation and misunderstanding. Ok, the conversation was about how Novamente could recursively self-improve itself into a runaway hard takeoff scenario. You're claiming that it can do so without the knowledge or skills of a competent programmer, with the very convincing argument Wrong. Care to elaborate at all? Or is your only purpose to communicate your slight irritation? - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=53154037-90851f
RE: [agi] RSI
Edward W. Porter writes: As I say, what is, and is not, RSI would appear to be a matter of definition. But so far the several people who have gotten back to me, including yourself, seem to take the position that that is not the type of recursive self improvement they consider to be RSI. Some people have drawn the line at coding. RSI they say includes modifying ones own code, but code of course is a relative concept, since code can come in higher and higher level languages and it is not clear where the distinction between code and non-code lies. As I had included comments along these lines in a previous conversation, I would like to clarify. That conversation was not specifically about a definition of RSI, it had to do with putting restrictions on the type of RSI we might consider prudent, in terms of cutting the risk of creating intelligent entities whose abilities grow faster than we can handle. One way to think about that problem is to consider that building an AGI involves taking a theory of mind and embodying it in a particular computational substrate, using one or more layers of abstraction built on the primitive operations of the substrate. That implementation is not the same thing as the mind model, it is one expression of the mind model. If we do not give arbitrary access to the mind model itself or its implementation, it seems safer than if we do -- this limits the extent that RSI is possible: the efficiency of the model implementation and the capabilities of the model do not change. Those capabilities might of course still be larger than was expected, so it is not a safety guarantee; further analysis using the particulars of the model and implementation, should be considered also. RSI in the sense of learning to learn better or learning to think better within a particular theory of mind seems necessary for any practical AGI effort so we don't have to code the details of every cognitive capability from scratch. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=49450400-3b2d82
RE: [agi] RSI
I wrote: If we do not give arbitrary access to the mind model itself or its implementation, it seems safer than if we do -- this limits the extent that RSI is possible: the efficiency of the model implementation and the capabilities of the model do not change. An obvious objection to this is that if the capabilities of the model include the ability to simulate a turing machine then the capabilities actually include everything computable. However, this issue being addressed is a practical one referring to what actually happens, and there are enormous practical issues involving resource limits of processing time and memory space that should be considered. Such consideration is part of a model-specific safety analysis. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=49457514-fe026f
RE: [agi] Religion-free technical content
Richard Loosemore: a) the most likely sources of AI are corporate or military labs, and not just US ones. No friendly AI here, but profit-making and mission-performing AI. Main assumption built into this statement: that it is possible to build an AI capable of doing anything except dribble into its wheaties, using the techiques currently being used. I have explained elsewhere why this is not going to work. If your explanations are convincing, smart people in industry and the military might just absorb them and then they still have more money and manpower than you do. When the first AGI is built, its first actions will be to make sure that nobody is trying to build a dangerous, unfriendly AGI. I often see it assumed that the step between first AGI is built (which I interpret as a functoning model showing some degree of generally-intelligent behavior) and god-like powers dominating the planet is a short one. Is that really likely? - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=48961147-810f59
RE: [agi] Religion-free technical content
Richard Loosemore writes: You must remember that the complexity is not a massive part of the system, just a small-but-indispensible part. I think this sometimes causes confusion: did you think that I meant that the whole thing would be so opaque that I could not understand *anything* about the behavior of the system? Like, all the characteristics of the system would be one huge emergent property, with us having no idea about where the intelligence came from? No, certainly not. I think the confusion here involves the distinction between Friendliness with a capital F (meaning a formal theory of what the term means and an intelligent system built to provably maintain that property in the mathematical, not verbal, sense), and friendliness with a lower case f, which relies on more human types of reasoning. It sounds to me like on the one hand you are saying that your system is complex and yet its behavior is not complex (at least in a particular but quite broad way - friendliness), as if you can bottle up the complexity in such a way that it has no important actual effects. In particular, when you write: You can build such a motivational system by controlling the system's agenda by diffuse connections into the thinking component that controls what it wants to do. It seems like doing that requires a rigorous understanding of the dynamics of the thinking component and I don't quite get how that can work in a guaranteed way since elements of the thinking component may change their nature in unpredictable ways and new elements of the thinking component may arise. If the nature of all thinking component elements are indeed perfectly predictable so that is impossible, I don't get where the complexity is. In your AGIRI paper, you seem to be saying that intelligence itself is an emergent property of a complex system (sorry to use words that everybody is probably allergic to), which seems to imply a global impact of the underlying complexity. I think (probably incorrectly) that I have a rough idea of how you intend to guide this complex system, and in fact I think that is likely the only way to go about it. It makes me nervous a bit when phrasings that bring capital-F Friendliness to mind are applied to designs that cannot possibly exhibit it. Basically, it rings alarm bells about overconfidence. However, the details of what you actually are working on will explain more than this conversation can, I'm sure. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=48433850-805862
RE: [agi] Religion-free technical content
Edward W. Porter writes: To Matt Mahoney. Your 9/30/2007 8:36 PM post referred to mine in reply to Derek Zahn and implied RSI (which I assume from context is a reference to Recursive Self Improvement) is necessary for general intelligence. So could you, or someone, please define exactly what its meaning is? Thanks for asking this question, I was just going to do so myself. If I am generally intelligent then I must be able to recursively self improve, even though all I can do is change some parameters in a particular neural structure. But that's just learning (even if it is learning better ways to learn). I don't think that sort of learning is what gives Singularitarians nightmares about AGI going BOOM a few days after birth, so defining what we mean and what we're actually afraid of could be very useful. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=48437261-c8a336
RE: [agi] Religion-free technical content
Richard: You agree that if we could get such a connection between the probabilities, we are home and dry? That we need not care about proving the friendliness if we can show that the probability is simply too low to be plausible? Yes, although the probability itself would have to be proven from first principles to be as strong as Friendliness. For any actual system such rigor seems as unlikely as Friendliness itself. Once the system is set up to behave according to a diffuse set of checks and balances (tens of thousands of ideas about what is right, rather than one single directive), it can never wander far from that set of constraints without noticing the departure immediately. Would you agree that IF such a design were feasible, you would not be able to think of any way to bollix it? * They have to be the right set of checks and balances, that completely cover ill-defined territory * Nothing unforseen can arise that is not covered by the designed-in checks and balances * The meaning of the constraints has to be applicable to all future developments somehow (e.g. the changing nature of humanity) * The meaning of the constraints and the complex items they operate on has to be immune to drift Given all that, nothing springs immediately into my little mind to disagree with your conclusion. Note that I think this type of approach is an excellent way to try for little-f friendliness, which is probably our best and only option. I like it a lot. Date: Mon, 1 Oct 2007 11:34:09 -0400 From: [EMAIL PROTECTED] To: agi@v2.listbox.com Subject: Re: [agi] Religion-free technical content Derek Zahn wrote: Richard Loosemore writes: You must remember that the complexity is not a massive part of the system, just a small-but-indispensible part. I think this sometimes causes confusion: did you think that I meant that the whole thing would be so opaque that I could not understand *anything* about the behavior of the system? Like, all the characteristics of the system would be one huge emergent property, with us having no idea about where the intelligence came from?No, certainly not. I think the confusion here involves the distinction between Friendliness with a capital F (meaning a formal theory of what the term means and an intelligent system built to provably maintain that property in the mathematical, not verbal, sense), and friendliness with a lower case f, which relies on more human types of reasoning. Derek, Your post raises several issues that I will try to get to in due course, but I want to deal with one of them quickly (if I can). I am attacking the very notion that there really is something that is mathematical Friendliness with a capital F, which can be proved formally rather than (something else). I am also stating that while this mythical provable-friendliness does not really exist (i.e. it will never be possible), there is something else that gives us exactly what we want, but is not a mathematical proof. Here is why. According to quantum mechanics there is a finite, non-zero probability that the Sun could suddenly quantum-tunnel itself to a new position inside the perfume department of Bloomingdales. There is no formal proof that it will not do this. There is no possibility of such a formal proof. But we accept that we do not need to worry about this happening because we have an idea of what the probability is. In essence, we know that for the Sun to do that, each atom in it would have to do the same thing all at once, and since the probability of each individual event is so small, and since they are all multiplied, the overall probability is stupidly small Now, of course I exaggerate for comedy, but the fact is that if you can make the event An AGI reneges on the motivations designed into it dependent on a very large number of improbable events all happening at once, then you can multipl the probabilities and come to a situation where the overall probability is vanishingly small. You agree that if we could get such a connection between the probabilities, we are home and dry? That we need not care about proving the friendliness if we can show that the probability is simply too low to be plausible? Right, now consider the nature of the design I propose: the motivational system never has an opportunity for a point failure: everything that happens is multiply-constrained (and on a massive scale: far more than is the case even in our own brains). Once the system is set up to behave according to a diffuse set of checks and balances (tens of thousands of ideas about what is right, rather than one single directive), it can never wander far from that set of constraints without noticing the departure immediately. Would you agree that IF such a design were feasible, you would not be able to think of any way
RE: [agi] Religion-free technical content
I suppose I'd like to see the list management weigh in on whether this type of talk belongs on this particular list or whether it is more appropriate for the singularity list. Assuming it's okay for now, especially if such talk has a technical focus: One thing that could improve safety is to reject the notion that AGI projects should be focused on, or even capable of, recursive self improvement in the sense of reprogramming its core implementation. Let's take Novamente as an example. Imagine that Ben G is able to take a break at some point from standing behind the counter of his Second Life pet store for a few years, and he gets his 1000-pc cluster and the implementation goes just as imagined and baby novamente is born some years down the road. At this point, Ben co. begin teaching it the difference between its virtual ass and a virtual hole in the ground. Novamente's model of mind is not the same thing as the C++ code that implements it; Baby Novamente has no particular affinity for computer programming or built in knowledge about software engineering. It cannot improve itself until the following things happen: 1) It acquires the knowledge and skills to become a competent programmer, a task that takes a human many years of directed training and practical experience. 2) It is given access to its own implementation and permission to alter it. 3) It understands its own implementation well enough to make a helpful change. Even if the years of time and effort were deliberately taken to make those things possible, further things would be necessary for it to be particularly worrisome: 1) Its programming abilities need to expand to the superhuman somehow -- a human equivalent programmer is not going to make radical improvements to a huge software system with man-decades of work behind it in a short period of time. A 100x or 1000x programming intelligence enhancement would be needed for that to happen. 2) The core implementation has to be incredibly flawed to squeeze orders of magnitude of extra efficiency into it. We're not really worried about a 30% improvement, we're worried about radical conceptual breakthroughs leading to huge peformance boosts. It stretches the imagination past its breaking point to imagine all of the above happening accidentally without Ben noticing. Therefore, to me, Novamente gets the Safe AGI seal of approval until such time as the above steps seem feasible and are undertaken. By that point, there will be years of time to consider its wisdom and hopefully apply some sort of friendliness theory to an actually dangerous stage. I think the development of such a theory is valuable (which is why I give money to SIAI) but I neither expect or want Ben to drop his research until it is ready. There is no need. I could imagine an approach to AGI that has at its core a reflexive understanding of its own implementation; a development pathway where algorithmic complexity theory, predictive models of its own code, code generation from an abstract specification language that forms a fluid self-model, unrestricted invention of new core components, and similar things. Such an approach might, in flights of imagination, be vulnerable to the oops, it's smarter than me now and I can't pull the plug scenario. But there's an easy answer to this: Don't build AGI that way. It is clearly not necessary for general intelligence (I don't understand my neural substrate and cannot rewire it arbitrarily at will). Surely certain AGI efforts are more dangerous than others, and the opaqueness that Yudkowski writes about is, at this point, not the primary danger. However, in that context, I think that Novamente is, to an extent, opaque in the sense that its actions may not be reduceable to anything clear (call such things emergent if you like, or just complex). If I understand Loosemore's argument, he might say that AGI without this type of opaqueness is inherently impossible, which could mean that Friendly AI is impossible. Suppose that's true... what do we do then? Minimize risks, I suppose. Perhaps certain protocol issues could be developed and agreed to. As an example: 1. A method to determine whether a given project at a certain developmental stage is dangerous enough to require restrictions. It is conceivable, for example, that any genetic programming homework, corewars game, or random programming error could accidentally generate the 200-instruction key to intelligence that wreaks havok on the planet... but it's so unlikely that forcing all programming to occur in cement bunkers seems like overkill. 2. Precautions for dangerous programs, such as limits to network access, limits to control of physical devices, and various types of deadman and emergency power cutoffs. I think we're a while away from needing any of this, but agree that it is not too soon to start thinking about it and, as has been pointed out,
RE: [agi] Religion-free technical content
Richard Loosemore writes: It is much less opaque. I have argued that this is the ONLY way that I know of to ensure that AGI is done in a way that allows safety/friendliness to be guaranteed. I will have more to say about that tomorrow, when I hope to make an announcement. Cool. I'm sure I'm not the only one eager to see how you can guarantee (read: prove) such specific detailed things about the behaviors of a complex system. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=48327693-309579
RE: [agi] HOW TO CREATE THE BUZZ THAT BRINGS THE BUCKS
Don Detrich writes: AGI Will Be The Most Powerful Technology In Human History – In Fact, So Powerful that it Threatens Us Admittedly there are many possible dangers with future AGI technology. We can think of a million horror stories and in all probability some of the problems that will crop up are things we didn’t anticipate. At this point it is pure conjecture. All new technologies have dangers, just like life in general. It'll be interesting to see if the horror stories about AGI follow the same pattern as they did for Nanotechnology... After many years and dollars of real nanotechnology research, the simplistic vision of the lone wolf researcher stumbling on a runaway self-replicator that turns the planet into gray goo became much more complicated and unlikely. Plus you can only write about gray goo for so long before it gets boring. Not to say that AGI is necessarily the same as Nanotechnology in its actual risks, or even that gray goo is less of an actual risk than writers speculated about, but it will be interesting to see if the scenario of a runaway self-reprogramming AI becomes similarly passe. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=47965814-e6c819
RE: [agi] Selfish promotion of AGI
Responding to Edward W. Porter: Thanks for the excellent message! I am perhaps too interested in seeing what the best response from the field of AGI might be to intelligent critics, and probably think of too many conversations in those terms; I did not mean to attack or criticise your statements, just guess at objections that a skeptic might make. Much of what you wrote here could be used in such a response. The truth is we really don’t know how big a good easy to computer representation of human level world knowledge would be. Right, or processing power to manipulate it. There are reasons to suspect it might be less than would be required for a molecular-level brain simulation. Intuitions about how much less will vary from individual to individual. 2. The software problem is solved. Ben Goertzel has solved it. I think most people will want more demonstration than a book review. I do too. Just to be clear, I'm not sure myself about whether Novamente has solved the software problem. It necessarily contains a large number of complex representational and implementation choices which I do not understand well enough to judge in an informed way. All I was trying to communicate is: prior to impressive demonstrations, nobody will believe that the software problem is solved. AI buzz has not been steady for 50 years. Except in SciFi, it has largely been missing in action for the last twenty years, since the overstated promises of the expert systems boom of the mid-eighties fell flat. If buzz means academic respectability, government grant levels, and availibility of risk capital, you're certainly right! I'm not sure what makes any of those groups tick so I am not sure what sort of buzzers would be effective. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=47484044-5944f6
RE: [agi] NVIDIA GPU's
Ben Goertzel writes: http://www.nvidia.com/page/home.html Anyone know what are the weaknesses of these GPU's as opposed to ordinary processors? They are good at linear algebra and number crunching, obviously. Is there some reason they would be bad at, say, MOSES learning? These parallel hardware innovations are indeed very exciting. I recently purchased a PC with two of these GPUs in it to play with. Like JoSH, I think that number crunching is The Way To Go. Unfortunately, these will be spectacularly bad at evaluating individuals for genetic programming. First, although they can do standard logic, program flow, and integer operations, that doesn't make very good use of the transistor count since the bulk of the silicon is dedicated to floating point arithmetic. Second, and more important, the programming model is SIMD, which means that the processors have to be running the same program. If, for example, and if statement's condition is satisfied on one processor but not the others, the others have to wait for the code inside to finish so they can all synchronize again. That would be terrible for evaluating heterogenous program trees. You're going to get your speedup over the coming years on that task from multicore CPUS that can run heterogenous threads. However, intuitively I think this massively parallel SIMD type of hardware might work rather well for propagation through your Probabilistic Logic Networks, depending on the details. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] NVIDIA GPU's
Moshe Looks writes: This is not quite correct; it really depends on the complexity of the programs one is evolving and the structure of the fitness function. For simple cases, it can really rock; see http://www.cs.ucl.ac.uk/staff/W.Langdon/ That's interesting work, thanks for the link! It's not immediately obvious, but the particular example there is a population of programs that estimate pi with up to 8 leaves from an alphabet of six tokens (2 constants and 4 basic arithmetic operations). The strategy used for parallelization is to run all programs currently waiting on a '+' operation, then run all programs doing a '-' operation and so on. If there are N operations (4 in this case), the population runs at 1/N speed (since the SIMD nature of the thing makes the others wait). So you're right, for simple cases like this one it only wastes 75% of the available processing power. It doesn't seem like it will scale very well though. Even on this simple task I bet a quad-core cpu is competitive with the GPU hardware. It does point out though that some things that are not intuitively data parallel can be executed effectively on a GPU. Do you personally think MOSES will run well on a GPU? - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] Another attempt to define General Intelligence, and some AGI design thoughts.
Robert Wensman writes: Has there been any work done previously in statistical, example driven deduction? Yes. In this AGI community, Pei Wang's NARS system is exactly that: http://nars.wang.googlepages.com/ Also, Ben Goertzel (et. al.) is building a system called Novamente (www.novamente.net) that has a Bayesian system of Probabilistic Logic Networks as its major representational scheme. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] Another attempt to define General Intelligence, and some AGI design thoughts.
Robert Wensman writes: Databases: 1. Facts: Contains sensory data records, and actuator records. 2. Theory: Contains memeplexes that tries to model the world. I don't usually think of 'memes' as having a primary purpose of modeling the world... it seems to me like the key to your whole approach is how you represent them (the schema of database 2). Could you elaborate a bit on that? - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] poll: what do you look for when joining an AGI group?
9. a particular AGI theoryThat is, one that convinces me it's on the right track. Now that you have run this poll, what did you learn from the responses and how are you using this information in your effort? - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] Symbol Grounding
I think probably AGI-curious person has intuitions about this subject. Here are mine: Some people, especially those espousing a modular software-engineering type of approach seem to think that a perceptual system basically should spit out a token for chair when it sees a chair, and then a reasoning system can take over to reason about chairs and what you might do with them -- and further it is thought that the reasoning about chairs part is really the essence of intelligence, whereas chair detection is just discardable pre-processing. My personal intuition says that by the time you have taken experience and boiled it down to a token labeled chair you have discarded almost everything important about the experience and all that is left is something that can be used by our logical inference systems. And although that ability to do logical inference (probabilistic or pure) is a super-cool thing that humans can do, it is a fairly minor part of our intelligence. Often I see AGI types referring to physical embodiment as a costly sideshow or as something that would be nice if a team of roboticists were available. But really, a simple robot is trivial to build, and even a camera on a pan/tilt base pointed at an interesting physical location is way easier to build than a detailed simulation world. The next objection is that image processing is too expensive and difficult. I guess my only thought about that it doesn't inspire confidence in an approach if the very first layer of neural processing is too hard. I suspect the real issue is that even if you do the image processing, then what? What do you do with the output? Ignoring those issues -- inventing a way of representing and manipulating knowledge, and assuming that sensory processes can create those data structures if built properly -- can work IF it turns out that brains are just really really bad at being intelligent. That is, if the extreme tip of the evolutionary iceberg (some thousands of generations of lightly-populated species) finally stumbled on the fluid symbol-manipulating abilities that define intelligence, and the rest of the historical structures are only mildly more important than organs that pump blood -- if that's true, thinking about all this low-level grunk is a waste of time. I actually hope that it's true, but I doubt it. To the first people who had the ability to code our magical symbol processing abilities on a machine, it must have seemed like an exciting theory. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] Symbol Grounding
One last bit of rambling in addition to my last post: When I assert that almost everything important gets discarded while merely distilling an array of rod and cone firings into a symbol for chair, it's fair to ask exactly what that other stuff is. Alas, I believe it is fundamentally impossible to tell you! I have seen some people attempt to communicate it, perhaps with a phrase like the play of shadow on the angle of the chair arm whose texture reminds me of the bus seat on that day with Julie in Madrid and the scratch on the leg which might be wood or might be plastic, sort of cone-like taking part of the chair's weight... The problem with trying to evoke the complexity and associative nature of the perceptual experience with a phrase like that is that every symbolist can easily nod and think about how all that gets encoded in their symbolic representation, with its nodes for bus and leg and the encoded memory of past events. But actually, the stuff is not at the right level for communicating linguistically so the above type of description is a made-up sham, more misleading than revealing. To the extent I have a theory about all this stuff, it's this: animals, including our evolutionary forebears, have concepts much like we do. However, somewhere recently in our history, something happened that greatly magnified our ability to use language, reason logically, and form dizzyingly abstract concepts. I think it's likely that it was a single thing (or that these are aspects of the single thing) rather than postulating three different radical innovations occurring at once. I'm not sure what that thing was, but I'd guess the following analogy: Concepts formed in some part of the brain grew handles of some kind, which allows them to be manipulated in a flexible combinational way by some new or improved dynamic processing mechanism that is either unique to us or is maybe vastly expanded from the abilities of lower species. Symbolists see the handles and the way they get tugged around and abstract it into combinatorial logics and linguistic grammars, but it doesn't do any good to tug handles around unless they are attached to the huge gooey blobs of mind-stuff, which are NOT logical or linguistic in nature. I'm philosophically a bottom upper because I think the hard and interesting questions have to do with the nature of those gooey blobby concepts. Examples of people who are trying to deal with that issue are Hawkins with his Hierarchical Temporal Memory, and Josh with his Interpolating Associative Memory (though those models are quite different in detail). I don't have a model myself. I do like to follow you top downers though as you do amazing things! - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] Pure reason is a disease.
Matt Mahoney writes: Below is a program that can feel pain. It is a simulation of a programmable 2-input logic gate that you train using reinforcement conditioning. Is it ethical to compile and run this program? - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] Get your money where your mouth is
Josh writes: http://www.netflixprize.com Thanks for bringing this up! I had heard of it but forgot about it. While I read about other people's projects/theories and build a robot for my own project, this will be a fun way to refresh myself on statistical machine learning techniques and statistics. I downloaded the data and it looks pretty easy to work with. See ya on the leaderboard :) - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] about AGI designers
YKY writes: There're several reasons why AGI teams are fragmented and AGI designers don't want to join a consortium: A. believe that one's own AGI design is superior B. want to ensure that the global outcome of AGI is friendly C. want to get bigger financial rewards D. There are no consortiums to join. I see talk about joining Novamente, but are they hiring? It might be possible to volunteer to work on peripheral things like AGISIM, but I sort of doubt that Ben is eager to train volunteers on the AGI-type code itself. On average, the cost/benefit of that would probably be quite poor. I see that AdaptiveAI has an opening for a programmer. We don't talk about them much, probably because they have chosen not to make much information availableabout what they're up to, beyond Peter Voss's vague overview paper. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e
RE: [agi] Pure reason is a disease.
Mark Waser writes: BTW, with this definition of morality, I would argue that it is a very rare human that makes moral decisions any appreciable percent of the time Just a gentle suggestion: If you're planning to unveil a major AGI initiative next month, focus on that at the moment. This stuff you have been arguing lately is quite peripheral to what you have in mind, except perhaps for the business model but in that area I see little compromise on more than subtle technical points. As I have begun to re-attach myself to the issues of AGI I have become suspicious of the ability or wisdom of attaching important semantics to atomic tokens (as I suspect you are going to attempt to do, along with most approaches), but I'd dearly like to contribute to something I thought had a chance. This stuff, though, belongs on comp.ai.philosophy (which is to say, it belongs unread). - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415user_secret=e9e40a7e