Re: [agi] Books
On 6/11/07, J Storrs Hall, PhD [EMAIL PROTECTED] wrote: I'll try to answer this and Mike Tintner's question at the same time. The typical GOFAI engine over the past decades has had a layer structure something like this: Problem-specific assertions Inference engine/database Lisp on top of the machine and OS. Now it turns out that this is plenty to build a system that can configure VAX computers or do NLP at the level of Why did you put the red block on the blue one? or What is the capital of the largest country in North America? The problem is that this leaves your symbols as atomic tokens in a logic-like environment, whose meaning is determined entirely from above, i.e. solely by virtue of their placement in expressions (or equivalently, links to other symbols in a semantic network). These formulations of a top layer were largely built on introspection, as was logic (and the Turing machine!). So chances are that a reasonable top layer could be built like that -- but the underpinnings are something a lot more capable than token-expression pattern matching. there's a big gap between the top layer(s) as found in AI programs and the bottom layers as found in existing programming systems. This is what I call Formalist Float in the book. It's not that any existing level is wrong, but there aren't enough of them, so that the higher ones aren't being built on the right primitives in current systems. Word-level concepts in the mind are much more elastic and plastic than logic tokens. You can build a factory where everything is made top-down, constructed with full attention to all its details. But if you try to build a farm that way, you'll do a huge amount of work and not get much -- your crops and livestock have to grow for themselves (and it's still a huge amount of work!). I think that the intermediate levels in the brain are built of robotic body controllers, mechanism with a flavor much like cybernetics, simply because that's what evolution had to work with. That's my working assumption in my experiments, anyway. Hi Josh, You haven't explained how your layered approach works, but I think you correctly exposed the problem of representation with logic-tokens. My solution to this is not exactly layers, but I see this as a difference between organic and inorganic knowledgebases. In Cyc, for example, the facts you enter into the KB remain the exact same way with the logic-tokens that you chose to enter them with. This is what I call inorganic. In an organic KB the facts will be *assimilated* into the KB via truth maintenance (old-fashioned term) or belief revision or cognitive dissonance resolution, etc. I think that mechanism is at the core of AGI. YKY - 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] Books
On Thursday 14 June 2007 02:12:29 am Joshua Fox wrote: I don't want to join any herd -- perhaps I just want to figure out why there is no AGI herd yet; as much a sociological question as a scientific one. It's probably worth pointing out to this group that for the first 25 years of its history, mainstream AI *was* an AGI herd. The universal assumption behind research was that they were going to build a full-fledged, language-speaking, cognitively complete, intelligent mind. Through the 60s AI got princely funding from ARPA (this in an era when computers cost, as a rule of thumb, a dollar per byte of memory). The funding, and AI with it, took a left turn during the 70s, towards applications, so that by the 80s the field was mostly expert systems and later neural nets. But the Golden Age had culminated in systems like Shrdlu and AM/Eurisko. If they had kept going on the original track, doing basic, general-intelligence-oriented research, they would have AGI now, I think. As far as I know, nobody even tried to put Shrdlu and Eurisko together to see what they would get until Novamente, which, I'm just guessing, is essentially that with a dollop of Copycat thrown in. (¿Ben?) Josh - 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] Books
Josh, Your point about layering makes perfect sense. I just ordered your book, but, impatient as I am, could I ask a question about this, though I've asked a similar question before: Why have not the elite of intelligent and open-minded leading AI researchers not attempted a multi-layered approach? Joshua 2007/6/10, J Storrs Hall, PhD [EMAIL PROTECTED]: Here's a big one: Levels of abstraction. I assume many of you are using a GUI mail client to read this. You're interacting with it in terms of windows, panels, boxes, buttons, menus, dragging and dropping. The GUI was written in terms of a toolkit that implements those concepts on top of an ontology involving events, queues, processes, locks, mutexes, and so forth. The program using the toolkit uses other libraries that are about rfc822-format messages, mime extensions, POP mailboxes, and the like. Typically, the programs and many of the libraries are written in programming languages which offer a model providing concepts like objects, methods, and functions. These in turn are based on lower-level languages where records, pointers, and memory allocation are the order of the day. In order to write code in any of this you have to understand, at least implicitly, the syntax of the language and use the translator that reads your code and compiles it into some internal form, using (most likely) an automatically generated shift-reduce parser. At some stage further down, the result will be assembly language for the machine you're running on, and then binary machine language. (And note that I somehow managed to leave out the entire level of the OS and hardware drivers and interrupt-level programming). There's just a big a stack of abstractions standing between the machine language and the transistors, in the machine architecture. Most AI (including a lot of what gets talked about here) is the equivalent of trying to implement the mail-reader directly in machine code (or transistors, for connectionists). Why people can't get the notion that the brain is going to be at least as ontologically deep as a desktop GUI is beyond me, but it's pretty much universal. Josh - 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] Books
I'll try to answer this and Mike Tintner's question at the same time. The typical GOFAI engine over the past decades has had a layer structure something like this: Problem-specific assertions Inference engine/database Lisp on top of the machine and OS. Now it turns out that this is plenty to build a system that can configure VAX computers or do NLP at the level of Why did you put the red block on the blue one? or What is the capital of the largest country in North America? The problem is that this leaves your symbols as atomic tokens in a logic-like environment, whose meaning is determined entirely from above, i.e. solely by virtue of their placement in expressions (or equivalently, links to other symbols in a semantic network). These formulations of a top layer were largely built on introspection, as was logic (and the Turing machine!). So chances are that a reasonable top layer could be built like that -- but the underpinnings are something a lot more capable than token-expression pattern matching. there's a big gap between the top layer(s) as found in AI programs and the bottom layers as found in existing programming systems. This is what I call Formalist Float in the book. It's not that any existing level is wrong, but there aren't enough of them, so that the higher ones aren't being built on the right primitives in current systems. Word-level concepts in the mind are much more elastic and plastic than logic tokens. You can build a factory where everything is made top-down, constructed with full attention to all its details. But if you try to build a farm that way, you'll do a huge amount of work and not get much -- your crops and livestock have to grow for themselves (and it's still a huge amount of work!). I think that the intermediate levels in the brain are built of robotic body controllers, mechanism with a flavor much like cybernetics, simply because that's what evolution had to work with. That's my working assumption in my experiments, anyway. Josh On Monday 11 June 2007 04:41:13 am Joshua Fox wrote: Josh, Your point about layering makes perfect sense. I just ordered your book, but, impatient as I am, could I ask a question about this, though I've asked a similar question before: Why have not the elite of intelligent and open-minded leading AI researchers not attempted a multi-layered approach? Joshua - 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: Reasoning in natural language (was Re: [agi] Books)
Interesting points, but I believe you can get around alot of the problems with two additional factors, a. using either large quantities of quality text, (ie novels, newspapers) or similar texts like newspapers. b. using a interactive built in 'checker' system, assisted learning where the AI could consult with humans in a simple way. Using something like this, you could check The moon is a dog and see that it has a really low probabilty, and if something else was possibly untrue, it could ask a few humans, and poll for the answer Is the moon a dog? This should allow for a large amount of basic information to be quickly gathered, and of a fairly high quality. James Matt Mahoney [EMAIL PROTECTED] wrote: --- Charles D Hixson wrote: Mark Waser wrote: The problem of logical reasoning in natural language is a pattern recognition problem (like natural language recognition in general). For example: - Frogs are green. Kermit is a frog. Therefore Kermit is green. - Cities have tall buildings. New York is a city. Therefore New York has tall buildings. - Summers are hot. July is in the summer. Therefore July is hot. After many examples, you learn the pattern and you can solve novel logic problems of the same form. Repeat for many different patterns. Your built in assumptions make you think that. There are NO readily obvious patterns is the examples you gave except on obvious example of standard logical inference. Note: * In the first clause, the only repeating words are green and Kermit. Maybe I'd let you argue the plural of frog. * In the second clause, the only repeating words are tall buildings and New York. I'm not inclined to give you the plural of city. There is also the minor confusion that tall buildings and New York are multiple words. * In the third clause, the only repeating words are hot and July. Okay, you can argue summers. * Across sentences, I see a regularity between the first and the third of As are B. C is A. Therefore, C is B. Looks far more to me like you picked out one particular example of logical inference and called it pattern matching. I don't believe that your theory works for more than a few very small, toy examples. Further, even if it did work, there are so many patterns that approaching it this way would be computationally intractable without a lot of other smarts. It's worse than that. Frogs are green. is a generically true statement, that isn't true in most particular cases. E.g., some frogs are yellow, red, and black without any trace of green on them that I've noticed. Most frogs may be predominately green (e.g., leopard frogs are basically green, but with black spots. Worse, although Kermit is identified as a frog, Kermit is actually a cartoon character. As such, Kermit can be run over by a tank without being permanently damaged. This is not true of actual frogs. OTOH, there *IS* a pattern matching going on. It's just not evident at the level of structure (or rather only partially evident). Were I to rephrase the sentences more exactly they would go something like this: Kermit is a representation of a frog. Frogs are typically thought of as being green. Therefore, Kermit will be displayed as largely greenish in overall hue, to enhance the representation. Note that one *could* use similar logic to deduce that Miss Piggy is more than 10 times as tall as Kermit. This would be incorrect. Thus, what is being discussed here is not mandatory characteristics, but representational features selected to harmonize an image with both it's setting and internal symbolisms. As such, only artistically selected features are chosen to highlight, and other features are either suppressed, or overridden by other artistic choices. What is being created is a dreamscape rather than a realistic image. On to the second example. Here again one is building a dreamscape, selecting harmonious imagery. Note that it's quite possible to build a dreamscape city where there are not tall buildings...or only one. (Think of the Emerald City of Oz. Or for that matter of the Sunset District of San Francisco. Facing in many directions you can't see a single building more than two stories tall.) But it's also quite realistic to imagine tall buildings. By specifying tall buildings, one filters out a different set of harmonious city images. What these patterns do is enable one to filter out harmonious images, etc. from the databank of past experiences. These are all valid criticisms. They explain why logical reasoning in natural language is an unsolved problem. Obviously simple string matching won't work. The system must also recognize sentence structure, word associations,
Re: Reasoning in natural language (was Re: [agi] Books)
On 6/11/07, James Ratcliff [EMAIL PROTECTED] wrote: Interesting points, but I believe you can get around alot of the problems with two additional factors, a. using either large quantities of quality text, (ie novels, newspapers) or similar texts like newspapers. b. using a interactive built in 'checker' system, assisted learning where the AI could consult with humans in a simple way. I would hope that a candidate AGI would have the capability of emailing anyone who has ever talked with it. ex: After a few minutes' chat, the AI asks the human for their email in case there it has any follow up questions - the same way any human interviewer might. If 10 humans are asked the same question, the statistically oddball response can probably be ignored (or reduced in weight) to clarify the answer. - 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: Reasoning in natural language (was Re: [agi] Books)
Correct, but I don't believe that systems (like Cyc) are doing this type of Active learning now, and it would help to gather quality information and fact-check it. Cyc does have some interesting projects where it takes a proposed statment and when a engineer is working with it, will go out and do a text match search in Google to check the validity of a statement, so would do soemthing like google search the moon is a dog returning 1/4bill so very unlikely. This goes one step towards my thoughts, but of course the Internet as a whole is not a trusted source for quality information, and would need to use a more refined base. Also OpenMind Common Sense (site down) is a very interesting project which does some information gathering using humans who log into the system and check and input information. It produced some intersting results, though on a limited basis. James Mike Dougherty [EMAIL PROTECTED] wrote: On 6/11/07, James Ratcliff wrote: Interesting points, but I believe you can get around alot of the problems with two additional factors, a. using either large quantities of quality text, (ie novels, newspapers) or similar texts like newspapers. b. using a interactive built in 'checker' system, assisted learning where the AI could consult with humans in a simple way. I would hope that a candidate AGI would have the capability of emailing anyone who has ever talked with it. ex: After a few minutes' chat, the AI asks the human for their email in case there it has any follow up questions - the same way any human interviewer might. If 10 humans are asked the same question, the statistically oddball response can probably be ignored (or reduced in weight) to clarify the answer. - 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/?; ___ James Ratcliff - http://falazar.com Looking for something... - Be a PS3 game guru. Get your game face on with the latest PS3 news and previews at Yahoo! Games. - 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] Books
Josh, Thanks for that answer on the layering of mind. It's not that any existing level is wrong, but there aren't enough of them, so that the higher ones aren't being built on the right primitives in current systems. Word-level concepts in the mind are much more elastic and plastic than logic tokens. Could I ask also that you take a stab at a psychological/sociological question: Why have not the leading minds of AI (considering for this purpose only the true creative thinkers with status in the community, however small a fraction that may be) taken a sufficiently multi-layered, grounded approach up to now? Isn't the need for grounding and deep-layering obvious to the most open-minded and intelligent of researchers? Joshua - 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] Books
On Monday 11 June 2007 02:06:35 pm Joshua Fox wrote: ... Could I ask also that you take a stab at a psychological/sociological question: Why have not the leading minds of AI (considering for this purpose only the true creative thinkers with status in the community, however small a fraction that may be) taken a sufficiently multi-layered, grounded approach up to now? Isn't the need for grounding and deep-layering obvious to the most open-minded and intelligent of researchers? Well, for one thing, the depth of the problem wasn't understood, and it a large extent one of the major contributions of the 50-year history of AI is to plumb it and give us a perspective. Today, AI along with cogsci and neuroscience, has given us a much better handle, I would venture to claim, on the scope of the problem. Second, it's not clear that the Newell Simon types won't ultimately be right. Our bodies are built around a flexible backbone, though no sane engineer would design an upright biped that way. We're built that way because we evolved from fish. There are probably plenty of backbones in the mind, which we can find more efficient replacements for once we've figured out how the whole business really works. But there will have to be a decade or two of experience with *working* AGI before it gets optimized to that point. Third, it's not clear that the top minds don't understand the problem perfectly well, but they have to work with what they've got, to advance the field to the point where they have something better. One could very reasonably characterize the last couple of decades' disregard of the goal of the integrated AI and instead the thrust of AI into COLT and modal logic and so forth as an attempt to build up the infrastructure -- and in fact it has been a very productive period, as far as the kinds of algorithms that are now available to build an AGI on are concerned. 1960s -- feet of clay 1970s -- legs of iron 1980s -- loins of bronze 1990s -- breast of brass 2000s -- head of silver Are we ready for the crown of gold yet? I like to think we're getting close :-) Josh - 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: Reasoning in natural language (was Re: [agi] Books)
--- James Ratcliff [EMAIL PROTECTED] wrote: Interesting points, but I believe you can get around alot of the problems with two additional factors, a. using either large quantities of quality text, (ie novels, newspapers) or similar texts like newspapers. b. using a interactive built in 'checker' system, assisted learning where the AI could consult with humans in a simple way. But that is not the problem I am trying to get around. A system that learns to solve logical word problems should be trainable on text like: - A greeb is a floogle. All floogles are blorg. Therefore... simply because it is something the human brain can do. Using something like this, you could check The moon is a dog and see that it has a really low probabilty, and if something else was possibly untrue, it could ask a few humans, and poll for the answer Is the moon a dog? This should allow for a large amount of basic information to be quickly gathered, and of a fairly high quality. James Matt Mahoney [EMAIL PROTECTED] wrote: --- Charles D Hixson wrote: Mark Waser wrote: The problem of logical reasoning in natural language is a pattern recognition problem (like natural language recognition in general). For example: - Frogs are green. Kermit is a frog. Therefore Kermit is green. - Cities have tall buildings. New York is a city. Therefore New York has tall buildings. - Summers are hot. July is in the summer. Therefore July is hot. After many examples, you learn the pattern and you can solve novel logic problems of the same form. Repeat for many different patterns. Your built in assumptions make you think that. There are NO readily obvious patterns is the examples you gave except on obvious example of standard logical inference. Note: * In the first clause, the only repeating words are green and Kermit. Maybe I'd let you argue the plural of frog. * In the second clause, the only repeating words are tall buildings and New York. I'm not inclined to give you the plural of city. There is also the minor confusion that tall buildings and New York are multiple words. * In the third clause, the only repeating words are hot and July. Okay, you can argue summers. * Across sentences, I see a regularity between the first and the third of As are B. C is A. Therefore, C is B. Looks far more to me like you picked out one particular example of logical inference and called it pattern matching. I don't believe that your theory works for more than a few very small, toy examples. Further, even if it did work, there are so many patterns that approaching it this way would be computationally intractable without a lot of other smarts. It's worse than that. Frogs are green. is a generically true statement, that isn't true in most particular cases. E.g., some frogs are yellow, red, and black without any trace of green on them that I've noticed. Most frogs may be predominately green (e.g., leopard frogs are basically green, but with black spots. Worse, although Kermit is identified as a frog, Kermit is actually a cartoon character. As such, Kermit can be run over by a tank without being permanently damaged. This is not true of actual frogs. OTOH, there *IS* a pattern matching going on. It's just not evident at the level of structure (or rather only partially evident). Were I to rephrase the sentences more exactly they would go something like this: Kermit is a representation of a frog. Frogs are typically thought of as being green. Therefore, Kermit will be displayed as largely greenish in overall hue, to enhance the representation. Note that one *could* use similar logic to deduce that Miss Piggy is more than 10 times as tall as Kermit. This would be incorrect. Thus, what is being discussed here is not mandatory characteristics, but representational features selected to harmonize an image with both it's setting and internal symbolisms. As such, only artistically selected features are chosen to highlight, and other features are either suppressed, or overridden by other artistic choices. What is being created is a dreamscape rather than a realistic image. On to the second example. Here again one is building a dreamscape, selecting harmonious imagery. Note that it's quite possible to build a dreamscape city where there are not tall buildings...or only one. (Think of the Emerald City of Oz. Or for that matter of the Sunset District of San Francisco. Facing in many directions you can't see a single building more than two stories tall.) But it's also quite realistic to imagine tall buildings. By specifying tall buildings, one filters
Re: [agi] Books
Josh: If you want to understand why existing approaches to AI haven't worked, try Beyond AI by yours truly Any major point or points worth raising here? - 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] Books
Josh: If you want to understand why existing approaches to AI haven't worked, try Beyond AI by yours truly Any major point or points worth raising here? Yo, troll, If you're really interested, then go get the book and stop wasting bandwidth. If you had any clue about AGI, you'd realize that any decent explanation is going to *have* to require a decent amount of text (since AI researchers haven't been totally clueless and running down alleys that could be disproved with a mere hundred words) and that the best (and most efficient) way to transfer that explanation is for you to just go get the book. (And Amazon e-mailed me yesterday that they had just/finally shipped my copy -- so it *is* available now) - 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] Books
Here's a big one: Levels of abstraction. I assume many of you are using a GUI mail client to read this. You're interacting with it in terms of windows, panels, boxes, buttons, menus, dragging and dropping. The GUI was written in terms of a toolkit that implements those concepts on top of an ontology involving events, queues, processes, locks, mutexes, and so forth. The program using the toolkit uses other libraries that are about rfc822-format messages, mime extensions, POP mailboxes, and the like. Typically, the programs and many of the libraries are written in programming languages which offer a model providing concepts like objects, methods, and functions. These in turn are based on lower-level languages where records, pointers, and memory allocation are the order of the day. In order to write code in any of this you have to understand, at least implicitly, the syntax of the language and use the translator that reads your code and compiles it into some internal form, using (most likely) an automatically generated shift-reduce parser. At some stage further down, the result will be assembly language for the machine you're running on, and then binary machine language. (And note that I somehow managed to leave out the entire level of the OS and hardware drivers and interrupt-level programming). There's just a big a stack of abstractions standing between the machine language and the transistors, in the machine architecture. Most AI (including a lot of what gets talked about here) is the equivalent of trying to implement the mail-reader directly in machine code (or transistors, for connectionists). Why people can't get the notion that the brain is going to be at least as ontologically deep as a desktop GUI is beyond me, but it's pretty much universal. Josh On Sunday 10 June 2007 05:49:36 am Mike Tintner wrote: Josh: If you want to understand why existing approaches to AI haven't worked, try Beyond AI by yours truly Any major point or points worth raising here? - 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] Books
Josh: Most AI (including a lot of what gets talked about here) is the equivalent of trying to implement the mail-reader directly in machine code (or transistors, for connectionists). Why people can't get the notion that the brain is going to be at least as ontologically deep as a desktop GUI is beyond me, but it's pretty much universal. Josh, Are you talking about levels of instruction (about how to handle the data) or levels of representation - that are ignored by AI? (As you may remember, I'm interested in the latter, and believe the brain processes info. simultaneously on at least 3 levels of abstractness/ concreteness). And what for you is the worst example of AI ignoring these levels of abstraction? - 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] Books
The problem of logical reasoning in natural language is a pattern recognition problem (like natural language recognition in general). For example: - Frogs are green. Kermit is a frog. Therefore Kermit is green. - Cities have tall buildings. New York is a city. Therefore New York has tall buildings. - Summers are hot. July is in the summer. Therefore July is hot. After many examples, you learn the pattern and you can solve novel logic problems of the same form. Repeat for many different patterns. Your built in assumptions make you think that. There are NO readily obvious patterns is the examples you gave except on obvious example of standard logical inference. Note: a.. In the first clause, the only repeating words are green and Kermit. Maybe I'd let you argue the plural of frog. b.. In the second clause, the only repeating words are tall buildings and New York. I'm not inclined to give you the plural of city. There is also the minor confusion that tall buildings and New York are multiple words. c.. In the third clause, the only repeating words are hot and July. Okay, you can argue summers. d.. Across sentences, I see a regularity between the first and the third of As are B. C is A. Therefore, C is B. Looks far more to me like you picked out one particular example of logical inference and called it pattern matching. I don't believe that your theory works for more than a few very small, toy examples. Further, even if it did work, there are so many patterns that approaching it this way would be computationally intractable without a lot of other smarts. - 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] Books
Mark Waser wrote: The problem of logical reasoning in natural language is a pattern recognition problem (like natural language recognition in general). For example: - Frogs are green. Kermit is a frog. Therefore Kermit is green. - Cities have tall buildings. New York is a city. Therefore New York has tall buildings. - Summers are hot. July is in the summer. Therefore July is hot. After many examples, you learn the pattern and you can solve novel logic problems of the same form. Repeat for many different patterns. Your built in assumptions make you think that. There are NO readily obvious patterns is the examples you gave except on obvious example of standard logical inference. Note: * In the first clause, the only repeating words are green and Kermit. Maybe I'd let you argue the plural of frog. * In the second clause, the only repeating words are tall buildings and New York. I'm not inclined to give you the plural of city. There is also the minor confusion that tall buildings and New York are multiple words. * In the third clause, the only repeating words are hot and July. Okay, you can argue summers. * Across sentences, I see a regularity between the first and the third of As are B. C is A. Therefore, C is B. Looks far more to me like you picked out one particular example of logical inference and called it pattern matching. I don't believe that your theory works for more than a few very small, toy examples. Further, even if it did work, there are so many patterns that approaching it this way would be computationally intractable without a lot of other smarts. It's worse than that. Frogs are green. is a generically true statement, that isn't true in most particular cases. E.g., some frogs are yellow, red, and black without any trace of green on them that I've noticed. Most frogs may be predominately green (e.g., leopard frogs are basically green, but with black spots. Worse, although Kermit is identified as a frog, Kermit is actually a cartoon character. As such, Kermit can be run over by a tank without being permanently damaged. This is not true of actual frogs. OTOH, there *IS* a pattern matching going on. It's just not evident at the level of structure (or rather only partially evident). Were I to rephrase the sentences more exactly they would go something like this: Kermit is a representation of a frog. Frogs are typically thought of as being green. Therefore, Kermit will be displayed as largely greenish in overall hue, to enhance the representation. Note that one *could* use similar logic to deduce that Miss Piggy is more than 10 times as tall as Kermit. This would be incorrect. Thus, what is being discussed here is not mandatory characteristics, but representational features selected to harmonize an image with both it's setting and internal symbolisms. As such, only artistically selected features are chosen to highlight, and other features are either suppressed, or overridden by other artistic choices. What is being created is a dreamscape rather than a realistic image. On to the second example. Here again one is building a dreamscape, selecting harmonious imagery. Note that it's quite possible to build a dreamscape city where there are not tall buildings...or only one. (Think of the Emerald City of Oz. Or for that matter of the Sunset District of San Francisco. Facing in many directions you can't see a single building more than two stories tall.) But it's also quite realistic to imagine tall buildings. By specifying tall buildings, one filters out a different set of harmonious city images. What these patterns do is enable one to filter out harmonious images, etc. from the databank of past experiences. - 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] Books
I've ended up with the following list. What do you think? * Ming Li and Paul Vitanyi, An Introduction to Kolmogorov Complexity and Its Applications, Springer Verlag 1997 * Marcus Hutter, Universal Artificial Intelligence: Sequential Decisions Based On Algorithmic Probability, Springer Verlag 2004 * Vladimir Vapnik, Statistical Learning Theory, Wiley-Interscience 1998 * Pedro Larrañaga, José A. Lozano (Editors), Estimation of Distribution Algorithms: A New Tool for Evolutionary Computation, Springer 2001 * Ben Goertzel, Cassio Pennachin (Editors), Artificial General Intelligence (Cognitive Technologies), Springer 2007 * Pei Wang, Rigid Flexibility: The Logic of Intelligence, Springer 2006 * Ben Goertzel, Matt Ikle', Izabela Goertzel, Ari Heljakka Probabilistic Logic Networks, in preparation * Juyang Weng et al., SAIL and Dav Developmental Robot Projects: the Developmental Approach to Machine Intelligence, publication list * Ralf Herbrich, Learning Kernel Classifiers: Theory and Algorithms, MIT Press 2001 * Eric Baum, What is Thought?, MIT Press 2004 * Marvin Minsky, The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind, Simon Schuster 2006 * Ben Goertzel, The Hidden Pattern: A Patternist Philosophy of Mind, Brown Walker Press 2006 * Ronald Brachman, Hector Levesque, Knowledge Representation and Reasoning, Morgan Kaufmann 2004 * Peter Gärdenfors, Conceptual Spaces: The Geometry of Thought, MIT Press 2004 * Wayne D. Gray (Editor), Integrated Models of Cognitive Systems, Oxford University Press 2007 * Logica Universalis, Birkhäuser Basel, January 2007 - 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
Reasoning in natural language (was Re: [agi] Books)
--- Charles D Hixson [EMAIL PROTECTED] wrote: Mark Waser wrote: The problem of logical reasoning in natural language is a pattern recognition problem (like natural language recognition in general). For example: - Frogs are green. Kermit is a frog. Therefore Kermit is green. - Cities have tall buildings. New York is a city. Therefore New York has tall buildings. - Summers are hot. July is in the summer. Therefore July is hot. After many examples, you learn the pattern and you can solve novel logic problems of the same form. Repeat for many different patterns. Your built in assumptions make you think that. There are NO readily obvious patterns is the examples you gave except on obvious example of standard logical inference. Note: * In the first clause, the only repeating words are green and Kermit. Maybe I'd let you argue the plural of frog. * In the second clause, the only repeating words are tall buildings and New York. I'm not inclined to give you the plural of city. There is also the minor confusion that tall buildings and New York are multiple words. * In the third clause, the only repeating words are hot and July. Okay, you can argue summers. * Across sentences, I see a regularity between the first and the third of As are B. C is A. Therefore, C is B. Looks far more to me like you picked out one particular example of logical inference and called it pattern matching. I don't believe that your theory works for more than a few very small, toy examples. Further, even if it did work, there are so many patterns that approaching it this way would be computationally intractable without a lot of other smarts. It's worse than that. Frogs are green. is a generically true statement, that isn't true in most particular cases. E.g., some frogs are yellow, red, and black without any trace of green on them that I've noticed. Most frogs may be predominately green (e.g., leopard frogs are basically green, but with black spots. Worse, although Kermit is identified as a frog, Kermit is actually a cartoon character. As such, Kermit can be run over by a tank without being permanently damaged. This is not true of actual frogs. OTOH, there *IS* a pattern matching going on. It's just not evident at the level of structure (or rather only partially evident). Were I to rephrase the sentences more exactly they would go something like this: Kermit is a representation of a frog. Frogs are typically thought of as being green. Therefore, Kermit will be displayed as largely greenish in overall hue, to enhance the representation. Note that one *could* use similar logic to deduce that Miss Piggy is more than 10 times as tall as Kermit. This would be incorrect. Thus, what is being discussed here is not mandatory characteristics, but representational features selected to harmonize an image with both it's setting and internal symbolisms. As such, only artistically selected features are chosen to highlight, and other features are either suppressed, or overridden by other artistic choices. What is being created is a dreamscape rather than a realistic image. On to the second example. Here again one is building a dreamscape, selecting harmonious imagery. Note that it's quite possible to build a dreamscape city where there are not tall buildings...or only one. (Think of the Emerald City of Oz. Or for that matter of the Sunset District of San Francisco. Facing in many directions you can't see a single building more than two stories tall.) But it's also quite realistic to imagine tall buildings. By specifying tall buildings, one filters out a different set of harmonious city images. What these patterns do is enable one to filter out harmonious images, etc. from the databank of past experiences. These are all valid criticisms. They explain why logical reasoning in natural language is an unsolved problem. Obviously simple string matching won't work. The system must also recognize sentence structure, word associations, different word forms, etc. Doing this requires a lot of knowledge about language and about the world. After those patterns are learned (and there are hundreds of thousands of them), then it will be possible to learn the more complex patterns associated with reasoning. The other criticism is that the statements are not precisely true. (July is cold in Australia). But the logic is still valid. It should be possible to train a purely logical system on examples using obviously false statements, like: - The moon is a dog. All dogs are made of green cheese. Therefore the moon is made of green cheese. The reasoning is correct, but confusing to many people. This fact argues (to me anyway) that logical
Re: [agi] Books
On 6/9/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: I've ended up with the following list. What do you think? I would like to add Locus Solum by Girard to this list, and then is seems to collapse into a black hole... Don't care? * Ming Li and Paul Vitanyi, An Introduction to Kolmogorov Complexity and Its Applications, Springer Verlag 1997 * Marcus Hutter, Universal Artificial Intelligence: Sequential Decisions Based On Algorithmic Probability, Springer Verlag 2004 * Vladimir Vapnik, Statistical Learning Theory, Wiley-Interscience 1998 * Pedro Larrañaga, José A. Lozano (Editors), Estimation of Distribution Algorithms: A New Tool for Evolutionary Computation, Springer 2001 * Ben Goertzel, Cassio Pennachin (Editors), Artificial General Intelligence (Cognitive Technologies), Springer 2007 * Pei Wang, Rigid Flexibility: The Logic of Intelligence, Springer 2006 * Ben Goertzel, Matt Ikle', Izabela Goertzel, Ari Heljakka Probabilistic Logic Networks, in preparation * Juyang Weng et al., SAIL and Dav Developmental Robot Projects: the Developmental Approach to Machine Intelligence, publication list * Ralf Herbrich, Learning Kernel Classifiers: Theory and Algorithms, MIT Press 2001 * Eric Baum, What is Thought?, MIT Press 2004 * Marvin Minsky, The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind, Simon Schuster 2006 * Ben Goertzel, The Hidden Pattern: A Patternist Philosophy of Mind, Brown Walker Press 2006 * Ronald Brachman, Hector Levesque, Knowledge Representation and Reasoning, Morgan Kaufmann 2004 * Peter Gärdenfors, Conceptual Spaces: The Geometry of Thought, MIT Press 2004 * Wayne D. Gray (Editor), Integrated Models of Cognitive Systems, Oxford University Press 2007 * Logica Universalis, Birkhäuser Basel, January 2007 - 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] Books
On 6/9/07, YKY (Yan King Yin) [EMAIL PROTECTED] wrote: I'm not aware of any book on pattern recognition with a view on AGI, except The Pattern Recognition Basis of Artificial Intelligence by Don Tveter (1998): http://www.dontveter.com/basisofai/basisofai.html You may look at The Cambridge Hankbook of Thinking and Reasoning first, especially the chapters on similarity and analogy. Thanks, it's interesting. - 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] Books
On 6/7/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: Reasoning about Uncertainty (Paperback) by Joseph Y. Halpern BTW, the .chm version of this book can be easily obtained on the net, as are many others you listed... I also recommand J Pearl's 2 books (Probabilistic Reasoning and Causality). Pattern Recognition, Third Edition (Hardcover) by Sergios Theodoridis (Author), Konstantinos Koutroumbas (Author) I have this one too, but the question is, how to apply pattern recognition in a logic-based setting? Knowledge Representation and Reasoning (The Morgan Kaufmann Series in Artificial Intelligence) by Ronald Brachman (Author), Hector Levesque (Author) A very good intro for anyone interested in logic-based AI. Two of the main points are: don't reinvent the wheel of KR; the tradeoff between KR expressiveness and efficiency of inference. Learning Kernel Classifiers: Theory and Algorithms (Adaptive Computation and Machine Learning) (Hardcover) by Ralf Herbrich (Author) I don't know how kernel methods can be applied in a logic-based setting. The math level of this one is also quite beyond me. Conceptual Spaces: The Geometry of Thought (Bradford Books) (Paperback) by Peter Gärdenfors (Author) I forgot what this book was about, will check it out again. Did you know that Gardenfors is very influential in the logic-based belief revision theory, the AGM (G=him) postulates? I'm not aware of any book on pattern recognition with a view on AGI, except *The Pattern Recognition Basis of Artificial Intelligence* by Don Tveter (1998): http://www.dontveter.com/basisofai/basisofai.html You may look at *The Cambridge Hankbook of Thinking and Reasoning* first, especially the chapters on similarity and analogy. YKY - 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] Books
--- YKY (Yan King Yin) [EMAIL PROTECTED] wrote: On 6/7/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: Pattern Recognition, Third Edition (Hardcover) by Sergios Theodoridis (Author), Konstantinos Koutroumbas (Author) I have this one too, but the question is, how to apply pattern recognition in a logic-based setting? The problem of logical reasoning in natural language is a pattern recognition problem (like natural language recognition in general). For example: - Frogs are green. Kermit is a frog. Therefore Kermit is green. - Cities have tall buildings. New York is a city. Therefore New York has tall buildings. - Summers are hot. July is in the summer. Therefore July is hot. After many examples, you learn the pattern and you can solve novel logic problems of the same form. Repeat for many different patterns. -- Matt Mahoney, [EMAIL PROTECTED] - 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