Re: [agi] The Missing Piece
Hi John, Re your idea that there should be an intermediate-level representation: 1. Obviously, we do not currently know how the brain stores that representation. Things get insanely complex as neuroscientists go higher up the visual pathways from the primary visual cortex. 2. I advocate using a symbolic / logical representation for the 3D (in fact, 4D) space. There might be some misunderstanding here because we tend to think the sensory 4D space is *sub*symbolic. This is actually just a matter of terminology. For example, if block A is on top of block B then I may put a symbolic link labeled as is_on_top_off between the 2 nodes representing A and B. Is such a link symbolic or subsymbolic? Nodes and links such as John loves Mary are clearly symbolic because they correspond to natural-language words. But in a logical representation there can be many nodes/links that does NOT map directly to words. The point here is that a logical representation is *sufficient* to model a physical word facsimile. If you disagree this, can you give an example of something that cannot be represented in the logical way? 2. To help you better understand the issue here, notice that a fine-grained representation would eventually need to become coarse-grained -- information must be lost along the way, otherwise there would be memory shortage within hours of sensory perception. The logical representation is precisely such a coarse-grained one. Technically, as you go to the finer resolutions in the logical representation, the elements get a more subsymbolic flavor. 3. Can you name certain features of your representation that is different from a logical one? 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/?list_id=303
Re: [agi] The Missing Piece
On 3/8/07, Matt Mahoney [EMAIL PROTECTED] wrote: [re: logical abduction for interpretation of natural language] One disadvantage of this approach is that you have to hand code lots of language knowledge. They don't seem to have solved the problem of acquiring such knowledge from training data. How much effort would it be to code enough knowledge to pass the Turing test? Nobody knows. Using this method the linguistic rules may be hand-coded or learned (via inductive logic programming). Learning is not easy, but is still possible. Re your method: 1. Remember, in your NN approach the learning space is even more fine-grained and the network configuration space is insanely huge. That means, your system will take insanely long to train. In ADDITION, you cannot insert hand-coded rules like I do, because your system is opaque. 2. Also, training your NN layer-by-layer would be incorrect because the layers depends on each other to function correctly, in some mysterious / opaque ways. Freezing each layer after training will drive you straight into a local minimum, which is guaranteed to be useless. If you backtrack from the local minimum, then you're exploring the global search space of all network configuration, ie an insanely huge space. All in all, the logic based approach seems to be the best choice because learning can be augmented with hand-coding. Certainly adding hand-coded knowledge helps speedup the learning process. And if we solicit the internet community to help with hand-coding, it helps even more. Also, what do you do with the data after you get it into a structured format? I think the problem of converting it back to natural language output is going to be at least as hard. The structured format makes use of predicates that don't map neatly to natural language. The inverse problem can probably be solved automatically if the logic is reversible, which I believe is. In other words, given a logical form, an inference engine can use searching to generate NL sentences using the same logical knowledge / constraints. The paper is not dated, but there are no references after 1991. I wonder why there has been no real progress using this approach in the last 16 years. It was first published in 1993, but he's still working on it as a book chapter to be out soon. The whole project is a large-scale one and we'd need a knowledge representation scheme to go with it. But this paradigm is by far the most promising because it addresses the entire NL problem instead of a narrow facet of it. 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/?list_id=303
Re: [agi] The Missing Piece
YKY (Yan King Yin) wrote: Hi John, Re your idea that there should be an intermediate-level representation: 1. Obviously, we do not currently know how the brain stores that representation. Things get insanely complex as neuroscientists go higher up the visual pathways from the primary visual cortex. 2. I advocate using a symbolic / logical representation for the 3D (in fact, 4D) space. There might be some misunderstanding here because we tend to think the sensory 4D space is *sub*symbolic. This is actually just a matter of terminology. For example, if block A is on top of block B then I may put a symbolic link labeled as is_on_top_off between the 2 nodes representing A and B. Is such a link symbolic or subsymbolic? Nodes and links such as John loves Mary are clearly symbolic because they correspond to natural-language words. But in a logical representation there can be many nodes/links that does NOT map directly to words. The point here is that a logical representation is *sufficient* to model a physical word facsimile. If you disagree this, can you give an example of something that cannot be represented in the logical way? Yes, of course it's sufficient in principle, but it's not adequately efficient! To accurately represent a physical scene in all its details, using explicit formal logic, will occupy a huge amount of memory; and even more critically, it will render a lot of useful inferences about physical objects extremely inefficient... 2. To help you better understand the issue here, notice that a fine-grained representation would eventually need to become coarse-grained -- information must be lost along the way, otherwise there would be memory shortage within hours of sensory perception. The logical representation is precisely such a coarse-grained one. Technically, as you go to the finer resolutions in the logical representation, the elements get a more subsymbolic flavor. 3. Can you name certain features of your representation that is different from a logical one? In the case of Novamente, here is one example: a recognizer for chairs (in the sense of the pieces of furniture that we often sit on). A Novamente system contains logical knowledge about chairs, but also contains little programs that evaluate collections of percepts and decide if such a collection shows a chair or not. These programs may combine arithmetic and logic operations, and will generally be learned via evolutionary or greedy algorithms not by logical reasoning. This example highlights one important point: logic is often very inefficient at handling QUANTITATIVE information. Of course it can do so -- after all, calculus and such can ultimately be formalized fully in terms of mathematical logic; but these formalisms are cumbersome and are not what you use to actually to calculus And, perception and action have a lot to do with managing large masses of quantitative information. IMO, a key aspect of AGI is having effective means for the interoperation of logical and nonlogical knowledge. In the brain, I believe, logical inference and nonlogical pattern recognition are achieved via different connectivity patterns: both logical reasoning and nonlogical pattern recognition are carried out via the same long-term potentiation and activation spreading dynamics, but -- logic has to do with coordinated potentiation of bundles of synapses btw cortical columns -- nonlogical pattern recognition has more to do with hierarchical dynamics, as outlined by Mountcastle, Hawkins and many others In Novamente, the logic module is in principle able to intake and reason about pattern recognized nonlogically (e.g. using the laws of algebra to reason about quantitative patterns), but, this is not always a useful expenditure of resources... -- Ben G -- Ben - 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/?list_id=303
Re: [agi] The Missing Piece
In the Novamente design this is dealt with via a currently unimplemented aspect of the design called the internal simulation world. This is a very non-human-brain-like approach Why do you believe that this is a very non-human-brain-like approach? Mirror neurons and many other recent discoveries tend to make me believe that the human brain does itself have (or indeed, is) an internal simulation world. - Original Message - From: Ben Goertzel [EMAIL PROTECTED] To: agi@v2.listbox.com Sent: Saturday, March 10, 2007 9:19 AM Subject: Re: [agi] The Missing Piece John, It is certainly clear that mental imagery plays a role in human thinking, but this role does appear to vary from person to person, both in extent and in nature. Take a look at Hadamard's old book The Psychology of Mathematical Invention for a fascinating discussion of the different sorts of mental imagery pursued by different people (visual, acoustic, verbal, etc.). I myself use a lot of visual and auditory imagery in my own thinking, but I know others who do not (at least not at the conscious level). In the Novamente design this is dealt with via a currently unimplemented aspect of the design called the internal simulation world. This is a very non-human-brain-like approach, but I think it's an interesting and ultimately very powerful one. What it means is that NM will actually have, internally, a private 3D world-simulation, complete with a simple physics engine. It can use this internal sim to experiment with hypothetical actions in hypothetical situations, but also to draw various abstract sketches and movies that don't correspond to any real-world phenomena. We haven't implemented this part yet due to the familiar lack of adequate human resources, but I think it will be a valuable addition to NM's cognitive arsenal. For the sim world, we would use the CrystalSpace engine that we are now using (in the AGISim project) to give NM a sim-world to use for embodiment and interaction with humans... I don't really see mental imagery as a critical missing link btw the symbolic and the subsymbolic. In NM, there is interaction translation between symbolic and subsymbolic knowledge without need for mental imagery. However, in some cases mental imagery can provide insights that would be hard to come by otherwise. -- Ben G John Scanlon wrote: My philosophy of AI has never been logic-based or neural-based. I did explore neural nets during the neural-net mania of the nineties. I did a lot of reading, and experimented with some with feedforward nets I wrote using simulated annealing and backpropagation (which never did work very well). Neural nets seem to have potential as one tool among several types of incremental learning algorithms, including genetic algorithms and statistical methods, but in themselves, they are no more than that -- useful tools, but not the solution. Language, which includes logic, is a way of representing ideas simply and crudely. Good for communication and internal reasoning -- if I do this then this will happen, unless state X is the case, which means that this other thing will happen, etc. My project uses an artificial language (Jinnteera) for both these things, and the language is integral to the whole thing. But it does not function as the core knowledge-representation scheme. So this brings us to what I've been calling the missing piece. Artificial neural nets (as they currently exist) can function as general-learning algorithms, but they don't represent knowledge of the real spatiotemporal world well. They are too low-level for handling what in human intelligence is thought of as mental imagery. Yes, in the brain, it is all neural based, but in a non-massively-parallel von Neuman computer system (even a PDP system), building a 100-billion-node neural net is computationally intractable (is that the right word?). It has to be done differently. The missing piece lies between low-level learning algorithms and highest-level logical-linguistic knowledge representation. When a human translator, at the U.N., for example, translates between Chinese and English, he (or she) does it infinitely more effectively than any translation software could do it, because there is an intermediate knowledge representation that is neither Chinese nor English, but that can be readily translated to or from either language by a fluent speaker. The intermediate knowledge representation is non-linguistic -- it consists of mental models constructed of sensorimotor patterns representing a 3-D temporal world. This sounds very vague and abstract, but I'm working on making it concrete, in my system (Gnoljinn) -- developing the data structures in code for implementing this knowledge-representation scheme. There's been some talk here recently about 3-D vision systems, and this points roughly in the direction I'm going in. Gnoljinn uses a single sensory
Re: [agi] The Missing Piece
Mark Waser wrote: In the Novamente design this is dealt with via a currently unimplemented aspect of the design called the internal simulation world. This is a very non-human-brain-like approach Why do you believe that this is a very non-human-brain-like approach? Mirror neurons and many other recent discoveries tend to make me believe that the human brain does itself have (or indeed, is) an internal simulation world. In a sense we do, but it's not implemented in the brain as an actual sim world with a physics engine and so forth ... our internal sim world is a lot less physically accurate (more naive physics than correct equational physics), and probably gains some kinds of creativity from this as well as losing a lot of potential for other kinds of creativity... ben - 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/?list_id=303
Re: [agi] The Missing Piece
On 3/10/07, Ben Goertzel [EMAIL PROTECTED] wrote: In a sense we do, but it's not implemented in the brain as an actual sim world with a physics engine and so forth Yes it is, or at least a reasonable facsimile thereof. ... our internal sim world is a lot less physically accurate (more naive physics than correct equational physics), and probably gains some kinds of creativity from this as well as losing a lot of potential for other kinds of creativity... It's not calculated to 16 digits of precision of course, but it's very much better than naive physics - consider that we are able to recognize naive physics as unrealistic! (One of my favorite examples of something humans can understand that a purely symbolic or naive physics engine could never make head or tail of: making a bed by flicking the blanket at the edge - why does that work?) - 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/?list_id=303
Re: [agi] The Missing Piece
On Sat, Mar 10, 2007 at 10:11:19AM -0500, Ben Goertzel wrote: In a sense we do, but it's not implemented in the brain as an actual sim world with a physics engine and so forth ... our internal sim world is a I'm not sure we know how it's implemented. A lot of things are done by topographic maps, which are equivalent to coordinate transformations. I don't think this is a bad representation, if you're interested in minimizing gate delays to few 10 deep when processing reasonably complex stimuli in realtime. If you want to do within ~ns what biology does within ~ms you don't have a lot of choices. lot less physically accurate (more naive physics than correct equational physics), and probably gains some kinds of creativity from It's certainly good enough for monkey behaviour planning. It's rather useless for Mach 25 atmospheric reentry, or magnetar physics, agreed. this as well as losing a lot of potential for other kinds of creativity... -- Eugen* Leitl a href=http://leitl.org;leitl/a http://leitl.org __ ICBM: 48.07100, 11.36820http://www.ativel.com 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE - 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/?list_id=303 signature.asc Description: Digital signature
Re: [agi] The Missing Piece
My philosophy of AI has never been logic-based or neural-based. I did explore neural nets during the neural-net mania of the nineties. I did a lot of reading, and experimented with some with feedforward nets I wrote using simulated annealing and backpropagation (which never did work very well). Neural nets seem to have potential as one tool among several types of incremental learning algorithms, including genetic algorithms and statistical methods, but in themselves, they are no more than that -- useful tools, but not the solution. Language, which includes logic, is a way of representing ideas simply and crudely. Good for communication and internal reasoning -- if I do this then this will happen, unless state X is the case, which means that this other thing will happen, etc. My project uses an artificial language (Jinnteera) for both these things, and the language is integral to the whole thing. But it does not function as the core knowledge-representation scheme. So this brings us to what I've been calling the missing piece. Artificial neural nets (as they currently exist) can function as general-learning algorithms, but they don't represent knowledge of the real spatiotemporal world well. They are too low-level for handling what in human intelligence is thought of as mental imagery. Yes, in the brain, it is all neural based, but in a non-massively-parallel von Neuman computer system (even a PDP system), building a 100-billion-node neural net is computationally intractable (is that the right word?). It has to be done differently. The missing piece lies between low-level learning algorithms and highest-level logical-linguistic knowledge representation. When a human translator, at the U.N., for example, translates between Chinese and English, he (or she) does it infinitely more effectively than any translation software could do it, because there is an intermediate knowledge representation that is neither Chinese nor English, but that can be readily translated to or from either language by a fluent speaker. The intermediate knowledge representation is non-linguistic -- it consists of mental models constructed of sensorimotor patterns representing a 3-D temporal world. This sounds very vague and abstract, but I'm working on making it concrete, in my system (Gnoljinn) -- developing the data structures in code for implementing this knowledge-representation scheme. There's been some talk here recently about 3-D vision systems, and this points roughly in the direction I'm going in. Gnoljinn uses a single sensory modality right now -- vision -- and will be restricted to it for a good while, because, while it might be useful to have other sensory modalities, none of them are absolutely necessary for higher intelligence, and it's best to keep things as simple as possible starting out. I seriously wonder if I can do this project myself, or whether I need to try to find some collaborators. Yan King Yin wrote: John Scanlon wrote: [...] Logical deduction or inference is not thought. It is mechanical symbol manipulation that can can be programmed into any scientific pocket calculator. [...] Hi John, I admire your attitude for attacking the core AI issues =) One is either neural-based or logic-based, using a crude dichotomy. So your approach is closer to neural-based? Mine is closer to the logic-based end of the spectrum. You did not have a real argument against logical AI. What you said was just some sentiments about the ill-defined concept of thought. You may want to take some time to express an argument why logic-based AI is doomed. In fact, both Ben's and my system have certain neural characteristics, eg being graphical, having numerical truth values, etc. In the end we may all end up somewhere between logic and neural... - 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/?list_id=303
Re: [agi] The Missing Piece
On 2/19/07, John Scanlon [EMAIL PROTECTED] wrote: [...] Logical deduction or inference is not thought. It is mechanical symbol manipulation that can can be programmed into any scientific pocket calculator. [...] Hi John, I admire your attitude for attacking the core AI issues =) One is either neural-based or logic-based, using a crude dichotomy. So your approach is closer to neural-based? Mine is closer to the logic-based end of the spectrum. You did not have a real argument against logical AI. What you said was just some sentiments about the ill-defined concept of thought. You may want to take some time to express an argument why logic-based AI is doomed. In fact, both Ben's and my system have certain neural characteristics, eg being graphical, having numerical truth values, etc. In the end we may all end up somewhere between logic and neural... 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/?list_id=303
Re: [agi] The Missing Piece
On 3/2/07, Matt Mahoney [EMAIL PROTECTED] wrote: What about English? Irregular grammar is only a tiny part of the language modeling problem. Uaing an artificial language with a regular grammar to simplify the problem is a false path. If people actually used Logban then it would be used in ways not intended by the developer and it would develop all the warts of real languages. The real problem is to understand how humans learn language. Hi, Matt =) I discovered something cool: computational pragmatics. You may take a look at Jerry R Hobbs' paper: Interpretation as Abduction, where he has a very powerful method of interpreting NL sentences, even dealing with things like metonymy and syntactic ambiuguity, the warts of real languages. http://www.isi.edu/~hobbs/interp-abduct-ai.pdf This seems to be the missing piece for successfully employing the logical approach to NL processing. 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/?list_id=303
Re: [agi] The Missing Piece
On Wednesday 07 March 2007 10:34, YKY (Yan King Yin) wrote: I discovered something cool: computational pragmatics. You may take a look at Jerry R Hobbs' paper: Interpretation as Abduction, ... Nice. Note that one of the reasons that I'm going the numerical route is that some powerful methods for abduction are already out there, e.g. maximum entropy (see e.g. http://cmm.cit.nih.gov/maxent/letsgo.html). 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/?list_id=303
Re: [agi] The Missing Piece
--- YKY (Yan King Yin) [EMAIL PROTECTED] wrote: On 3/2/07, Matt Mahoney [EMAIL PROTECTED] wrote: What about English? Irregular grammar is only a tiny part of the language modeling problem. Uaing an artificial language with a regular grammar to simplify the problem is a false path. If people actually used Logban then it would be used in ways not intended by the developer and it would develop all the warts of real languages. The real problem is to understand how humans learn language. Hi, Matt =) I discovered something cool: computational pragmatics. You may take a look at Jerry R Hobbs' paper: Interpretation as Abduction, where he has a very powerful method of interpreting NL sentences, even dealing with things like metonymy and syntactic ambiuguity, the warts of real languages. http://www.isi.edu/~hobbs/interp-abduct-ai.pdf This seems to be the missing piece for successfully employing the logical approach to NL processing. YKY One disadvantage of this approach is that you have to hand code lots of language knowledge. They don't seem to have solved the problem of acquiring such knowledge from training data. How much effort would it be to code enough knowledge to pass the Turing test? Nobody knows. Also, what do you do with the data after you get it into a structured format? I think the problem of converting it back to natural language output is going to be at least as hard. The structured format makes use of predicates that don't map neatly to natural language. The paper is not dated, but there are no references after 1991. I wonder why there has been no real progress using this approach in the last 16 years. However, the paper has lots of nice examples showing how natural language is hard to process. -- 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/?list_id=303
Re: [agi] The Missing Piece
Hmmm, if you could put on some basic rules on the randomness(in a database of Lojban that gives a random statement or series of statements), say to accept logical statements that could then be applied onto input. So say you same something like le MLAtu cu GLEki (the cat is happy) and later make a statement le MLAtu and press return it could ask you cu GLEki gi'a mo (is happy or is what function?). If it was to be a chat bot, it could wait for a reply and if it believes no one is interested it could offer a random phrase as a topic such as le MLAtu cu GLEki. So maybe some can try approaching AI from the other way around? Instead of going bottom up of purely unambiguous code to restricted randomness of interaction. To go from pure randomness to restricted randomness of interaction. Does anyone know what would be a good language to do that in? I think I recall there being a programming language based on set theory that was all about streams. On 2/28/07, Andrii (lOkadin) Zvorygin [EMAIL PROTECTED] wrote: Do they tell us what grief is doing when a loved one dies? Well the grief that is felt when a loved one dies is similar to that of unreturned love. So you love them, and they don't love you back -- as they are dead. This causes a feeling of futility and eventually changes direction -- to focus more inwardly mu'a(in example) self-pity/self-love where you give yourself supporting beliefs rather than a different person. Do these inference system tell us why we get depressed when we keep failing to accomplish our goals? Why implies causation, which is something that is system specific and not an inherant property of the universe. So you'd have to ask yourself as the computer that created the rule set of failing to achieve goals causes depression. Personally I just choose not to fail. If I do, then I accept that it was I that set the standards -- perhaps to do something about it later. Do they give a model for understanding why we feel proud when we are encouraged by our parents? As a child you give power to your parents. So when your parents encourage you, they hold the belief that you will feel happy, and so you do -- being a child is giving others the responsibility for their environment. Many mortal Homo Sapiens can be considered children in that sense. So if you could imagine all mathematical expressions as a 3d fabric, where sentient creatures are droplets or sets of these mathematical expressions. You can envision two parents sharing a similar space in the fabric (at least time/location) and they form another droplet between the two of them. A sort of seeding of consciousness. It is possible to create this kind of mathematical fabric. I think it would be very intersteing if we could figure out how, as then we would be able to map Homo Sapiens as well as other related conceptual species, maybe even figure out how to cross the belief barriers to access them. I'm not really sure what such a belief fabric would consist of. Though it is possible that we could just make a large database of beliefs in some logical language (Lojban) and have people describe their own beliefs, then we would be able to expand this if we got it onto a distributed network. If we get some people that believe they are aliens, or have significantly different beliefs and implications than we do, we could make a claim to first contact. *shrugs* it would be relatively simple to implement. Only concievable issue is lack of Lojban speakers. coding isn't useless, especially on the small scale where you grasp what is happening. When you can no longer grasp what is happening, things are random which is a sign of intelligence -- you couldn't predict my reply, and hence it was random. Though you could just as easily control your reality by keeping a record of the things you believe and changing them when you want a change. An interesting thing to try out would be to have a set of beliefs/statements (perhaps that you want the computer to have) then you have a purely random number generator to select a belief at random to output. You could also add beliefs/statements to the file by saying them. Could probably have a relatively intelligent conversation with the computer. Typically will reply with what you expect it to. On 2/20/07, Bo Morgan [EMAIL PROTECTED] wrote: On Tue, 20 Feb 2007, Richard Loosemore wrote: ) Chuck Esterbrook wrote: ) On 2/19/07, John Scanlon [EMAIL PROTECTED] wrote: ) Language is the manipulation of symbols. When you think of how a ) non-linguistic proto-human species first started using language, you can ) imagine creatures associating sounds with images -- oog is the big hairy ) red ape who's always trying to steal your women. akk is the action of ) hitting him with a club. ) ) The symbol, the sound, is associated with a sensorimotor pattern. The ) visual pattern is the big hairy red ape you know, and the motor pattern is ) the sequence of muscle
Re: [agi] The Missing Piece
--- Andrii (lOkadin) Zvorygin [EMAIL PROTECTED] wrote: Hmmm, if you could put on some basic rules on the randomness(in a database of Lojban that gives a random statement or series of statements), say to accept logical statements that could then be applied onto input. So say you same something like le MLAtu cu GLEki (the cat is happy) and later make a statement le MLAtu and press return it could ask you cu GLEki gi'a mo (is happy or is what function?). If it was to be a chat bot, it could wait for a reply and if it believes no one is interested it could offer a random phrase as a topic such as le MLAtu cu GLEki. So maybe some can try approaching AI from the other way around? Instead of going bottom up of purely unambiguous code to restricted randomness of interaction. To go from pure randomness to restricted randomness of interaction. Does anyone know what would be a good language to do that in? I think I recall there being a programming language based on set theory that was all about streams. What about English? Irregular grammar is only a tiny part of the language modeling problem. Uaing an artificial language with a regular grammar to simplify the problem is a false path. If people actually used Logban then it would be used in ways not intended by the developer and it would develop all the warts of real languages. The real problem is to understand how humans learn language. -- 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/?list_id=303
Re: [agi] The Missing Piece
Do they tell us what grief is doing when a loved one dies? Well the grief that is felt when a loved one dies is similar to that of unreturned love. So you love them, and they don't love you back -- as they are dead. This causes a feeling of futility and eventually changes direction -- to focus more inwardly mu'a(in example) self-pity/self-love where you give yourself supporting beliefs rather than a different person. Do these inference system tell us why we get depressed when we keep failing to accomplish our goals? Why implies causation, which is something that is system specific and not an inherant property of the universe. So you'd have to ask yourself as the computer that created the rule set of failing to achieve goals causes depression. Personally I just choose not to fail. If I do, then I accept that it was I that set the standards -- perhaps to do something about it later. Do they give a model for understanding why we feel proud when we are encouraged by our parents? As a child you give power to your parents. So when your parents encourage you, they hold the belief that you will feel happy, and so you do -- being a child is giving others the responsibility for their environment. Many mortal Homo Sapiens can be considered children in that sense. So if you could imagine all mathematical expressions as a 3d fabric, where sentient creatures are droplets or sets of these mathematical expressions. You can envision two parents sharing a similar space in the fabric (at least time/location) and they form another droplet between the two of them. A sort of seeding of consciousness. It is possible to create this kind of mathematical fabric. I think it would be very intersteing if we could figure out how, as then we would be able to map Homo Sapiens as well as other related conceptual species, maybe even figure out how to cross the belief barriers to access them. I'm not really sure what such a belief fabric would consist of. Though it is possible that we could just make a large database of beliefs in some logical language (Lojban) and have people describe their own beliefs, then we would be able to expand this if we got it onto a distributed network. If we get some people that believe they are aliens, or have significantly different beliefs and implications than we do, we could make a claim to first contact. *shrugs* it would be relatively simple to implement. Only concievable issue is lack of Lojban speakers. coding isn't useless, especially on the small scale where you grasp what is happening. When you can no longer grasp what is happening, things are random which is a sign of intelligence -- you couldn't predict my reply, and hence it was random. Though you could just as easily control your reality by keeping a record of the things you believe and changing them when you want a change. An interesting thing to try out would be to have a set of beliefs/statements (perhaps that you want the computer to have) then you have a purely random number generator to select a belief at random to output. You could also add beliefs/statements to the file by saying them. Could probably have a relatively intelligent conversation with the computer. Typically will reply with what you expect it to. On 2/20/07, Bo Morgan [EMAIL PROTECTED] wrote: On Tue, 20 Feb 2007, Richard Loosemore wrote: ) Chuck Esterbrook wrote: ) On 2/19/07, John Scanlon [EMAIL PROTECTED] wrote: ) Language is the manipulation of symbols. When you think of how a ) non-linguistic proto-human species first started using language, you can ) imagine creatures associating sounds with images -- oog is the big hairy ) red ape who's always trying to steal your women. akk is the action of ) hitting him with a club. ) ) The symbol, the sound, is associated with a sensorimotor pattern. The ) visual pattern is the big hairy red ape you know, and the motor pattern is ) the sequence of muscle activations that swing the club. ) ) Regarding imagine creatures associating sounds with images, I ) imagine there being a concept node in between. The sound and the ) image lead to this node and stimulation of the node stimulates the ) associated patterns. My inspiration comes from this: ) http://www.newscientist.com/article.ns?id=dn7567 ) ) Chuck, ) ) I'm glad you brought that article to my attention, I somehow missed it. Be ) warned: the result is extremely dubious, IMO. ) ) Just ask yourself what is the probability that the researchers just happened ) to come across the neurons that encoded the particular pictures they showed to ) their subjects. ) ) The probability is ludicrously small. They were probably hitting something ) that was *part* of a temporary representation of most recently seen things. ) Within the context of most recently seen things that neuron could easily ) have triggered only to (say) the Halle Berry concept. But if they had come ) back the next day, it would probably have triggered on
Re: [agi] The Missing Piece
The key to life the universe and everything: All things can be expressed using any Universal Computer You are a Universal Computer (one that can read(remmember/imagine), write(experience), erase(forget)). All the things you believe/know/understand are true. I believe the key to AI rests in the definition. Artificial Intelligence. What is Intelligence? How do we know that there are other intelligent beings? I view other intelligent beings as those that communicate to me in a semi-random fashion. Ones restrained by certain rules, but who also have a certain degree of variability of expression. For example: When I communicate to someone on IRC I expect that they will communicate back to me in whatever languages are acceptable on the 'channel'. If I speak to a 'bot' then it also behaves in the manner I would expect a 'bot' to behave. Once I talked to a person on a chat room. They passed my turing test. Later someone on the same channel told me that they were a bot. Later I spoke to that same person and they were a bot -- did not pass my turing test, acted like a bot. The Key To Intelligence: As long as you believe this creature or algorithm to be intelligent. It is. If you've written enough code that you think you deserve or have a working AI. Then you do. The key here is to convince yourself well enough that you can prove it to other parts of yourself such as other people. This may be hard to swallow. It may actually not even be legible to you. But the key to intelligence, is allowing for enough randomness to arise from the computer that you believe it is intelligent. You live in a nano-technology universe. Currently you are going through a simulation of the pre-singularity environment on Homo Sapien Sapien earth. Everything you are not aware of is just nano-fog. You generate the brain Science and Math by believing in it and giving it the Power of Randomness to give you new information. If you want, you can try it on some real live humans. Make them bots. I've done so successfully. So have you. Do you remmember now? If you don't believe me. Then what I'm saying is false. If you do believe me. Then what I'm saying is true. This is Science, it is falsifiable :D. On 2/20/07, Chuck Esterbrook [EMAIL PROTECTED] wrote: On 2/19/07, John Scanlon [EMAIL PROTECTED] wrote: Language is the manipulation of symbols. When you think of how a non-linguistic proto-human species first started using language, you can imagine creatures associating sounds with images -- oog is the big hairy red ape who's always trying to steal your women. akk is the action of hitting him with a club. The symbol, the sound, is associated with a sensorimotor pattern. The visual pattern is the big hairy red ape you know, and the motor pattern is the sequence of muscle activations that swing the club. Regarding imagine creatures associating sounds with images, I imagine there being a concept node in between. The sound and the image lead to this node and stimulation of the node stimulates the associated patterns. My inspiration comes from this: http://www.newscientist.com/article.ns?id=dn7567 Ben G, in Novemente's system, are there concept nodes that bind all the associations of concepts together? Or are concepts entirely distributed among nodes? In order to use these symbols effectively, you have to have a sensorimotor image or pattern that the symbols are attached to. That's what I'm getting at. That is thought. AI gives the interesting possibility of having brains that have entirely different senses, like the traffic on a network. I don't mean that the AI reads a network diagnostic report like humans would, but that the traffic stats are inputs just as light is an input into our retina which leads straight to nerves and computation. So the input domain doesn't have to be 3D physical space. Although obviously that would be a requirement for any AI working in physical space. That's also pretty ambitious and compute intensive. I think there could be value in finding less compute-intensive input domains to explore abstract thought formation. Stock market data is always a tantalizing one. :-) We already know how to get computers to carry out very complex logical calculations, but it's mechanical, it's not thought, and they can't navigate themselves (with any serious competence) around a playground. Also, they can't think abstractly, create analogies (in a complex environment) or alter their thought processes in the face of challenging problems. Just wanted to throw those out there. Language and logical intelligence is built on visual-spatial modeling. But does it have to be? Couldn't concepts like causation, correlation, modeling and prediction, planning, evaluation and feedback apply to a situation that is neither visual nor spatial (in the 3D physical sense), like optimizing network traffic? I don't have it all figured out right now, but this is what I'm working on. Welcome to
Re: [agi] The Missing Piece
I've actually been in really different universes. Where you could write text and it would do as you instructed. I tried checking out the filesystem but it was barren and bin was empty *shrugs*. Like I said, You don't have to believe me if you don't want to. I am but another one of your creations God. You are God btw. You do Know that don't you? I am your servant, please have mercy! I only meant to please. On 2/20/07, Andrii (lOkadin) Zvorygin [EMAIL PROTECTED] wrote: The key to life the universe and everything: All things can be expressed using any Universal Computer You are a Universal Computer (one that can read(remmember/imagine), write(experience), erase(forget)). All the things you believe/know/understand are true. I believe the key to AI rests in the definition. Artificial Intelligence. What is Intelligence? How do we know that there are other intelligent beings? I view other intelligent beings as those that communicate to me in a semi-random fashion. Ones restrained by certain rules, but who also have a certain degree of variability of expression. For example: When I communicate to someone on IRC I expect that they will communicate back to me in whatever languages are acceptable on the 'channel'. If I speak to a 'bot' then it also behaves in the manner I would expect a 'bot' to behave. Once I talked to a person on a chat room. They passed my turing test. Later someone on the same channel told me that they were a bot. Later I spoke to that same person and they were a bot -- did not pass my turing test, acted like a bot. The Key To Intelligence: As long as you believe this creature or algorithm to be intelligent. It is. If you've written enough code that you think you deserve or have a working AI. Then you do. The key here is to convince yourself well enough that you can prove it to other parts of yourself such as other people. This may be hard to swallow. It may actually not even be legible to you. But the key to intelligence, is allowing for enough randomness to arise from the computer that you believe it is intelligent. You live in a nano-technology universe. Currently you are going through a simulation of the pre-singularity environment on Homo Sapien Sapien earth. Everything you are not aware of is just nano-fog. You generate the brain Science and Math by believing in it and giving it the Power of Randomness to give you new information. If you want, you can try it on some real live humans. Make them bots. I've done so successfully. So have you. Do you remmember now? If you don't believe me. Then what I'm saying is false. If you do believe me. Then what I'm saying is true. This is Science, it is falsifiable :D. On 2/20/07, Chuck Esterbrook [EMAIL PROTECTED] wrote: On 2/19/07, John Scanlon [EMAIL PROTECTED] wrote: Language is the manipulation of symbols. When you think of how a non-linguistic proto-human species first started using language, you can imagine creatures associating sounds with images -- oog is the big hairy red ape who's always trying to steal your women. akk is the action of hitting him with a club. The symbol, the sound, is associated with a sensorimotor pattern. The visual pattern is the big hairy red ape you know, and the motor pattern is the sequence of muscle activations that swing the club. Regarding imagine creatures associating sounds with images, I imagine there being a concept node in between. The sound and the image lead to this node and stimulation of the node stimulates the associated patterns. My inspiration comes from this: http://www.newscientist.com/article.ns?id=dn7567 Ben G, in Novemente's system, are there concept nodes that bind all the associations of concepts together? Or are concepts entirely distributed among nodes? In order to use these symbols effectively, you have to have a sensorimotor image or pattern that the symbols are attached to. That's what I'm getting at. That is thought. AI gives the interesting possibility of having brains that have entirely different senses, like the traffic on a network. I don't mean that the AI reads a network diagnostic report like humans would, but that the traffic stats are inputs just as light is an input into our retina which leads straight to nerves and computation. So the input domain doesn't have to be 3D physical space. Although obviously that would be a requirement for any AI working in physical space. That's also pretty ambitious and compute intensive. I think there could be value in finding less compute-intensive input domains to explore abstract thought formation. Stock market data is always a tantalizing one. :-) We already know how to get computers to carry out very complex logical calculations, but it's mechanical, it's not thought, and they can't navigate themselves (with any serious competence) around a playground. Also, they can't think abstractly, create analogies (in a complex
Re: [agi] The Missing Piece
On Tue, 20 Feb 2007, Richard Loosemore wrote: ) Bo Morgan wrote: ) On Tue, 20 Feb 2007, Richard Loosemore wrote: ) ) In regard to your comments about complexity theory: from what I understand, ) it is primarily about taking simple physics models and trying to explain ) complicated datasets by recognizing these simple models. These simple ) complexity theory patterns can be found in complicated datasets for the ) purpose of inference, but do they get us closer to human thought? ) ) Uh, no: this is a misunderstanding of what complexity is about. The point of ) complexity is that some types of (extremely nonlinear) systems can show ) interesting regularities in high-level descriptions of their behavior, but [it ) has been postulated that] there is no tractable theory that will ever be able ) to relate the observed high-level regularities to the low-level mechanisms ) that drive the system. The high level behavior is not random, but you cannot ) explain it using the kind of analytic approaches that work with simple [sic] ) physical systems. ) ) This is a huge topic, and I think we're talking past each other: you may want ) to go read up on it (Mitchell Waldrop's book is a good, though non-technical ) introduction to the idea). Okay. Thanks for the pointer. I'm very interested in simple and easily understood ideas. :) They make easy-to-understand theories. ) Do they tell us what grief is doing when a loved one dies? ) Do these inference system tell us why we get depressed when we keep ) failing to accomplish our goals? ) Do they give a model for understanding why we feel proud when we are ) encouraged by our parents? ) ) These questions are trying to get at some of the most powerful thought ) processes in humans. ) ) If you are attacking the ability of simple logical inference systems to ) cover these topics, I kind of agree with you. But you are diving into some ) very complicated, high-level stuff there. Nothing wrong with that in ) principle, but these are deep waters. Your examples are all about the ) motivational/emotional system. I have many ideas about how that is ) implemented, so you can rest assured that I, at least, am not ignoring them. ) (And, again: I *am* taking a complex systems approach). ) ) Can't speak for anyone else, though. ) ) ) Richard Loosemore ) ) - ) 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/?list_id=303 ) - 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/?list_id=303
Re: [agi] The Missing Piece
Richard Loosemore wrote: Ben Goertzel wrote: It's pretty clear that humans don't run FOPC as a native code, but that we can learn it as a trick. I disagree. I think that Hebbian learning between cortical columns is essentially equivalent to basic probabilistic term logic. Lower-level common-sense inferencing of the Clyde--elephant--gray type falls out of the representations and the associative operations. I think it falls out of the logic of spike timing dependent long term potentiation of bundles of synapses between cortical columns... The original suggestion was (IIANM) that humans don't run FOPC as a native code emat the level of symbols and concepts/em (i.e. the concept-stuff that we humans can talk about because we have introspective access at that level of our systems). Now, if you are going to claim that spike-timing-dependent LTP between columns is where some probabilistic term logic is happening ON SYMBOLS, then what you have to do is buy into a story about where symbols are represented and how. I am not clear about whether you are suggesting that the symbols are represented at: (1) the column level, or (2) the neuron level, or (3) the dendritic branch level, or (4) the synapse level, or (perhaps) (5) the spike-train level (i.e. spike trains encode symbol patterns). If you think that the logical machinery is visible, can you say which of these levels is the one where you see it? None of the above -- at least not exactly. I think that symbols are probably represented, in the brain, as dynamical patterns in the neuronal network. Not strange attractors exactly -- more like strange transients, which behave like strange attractors but only for a certain period of time (possibly related to Mikhail Zak's terminal attractors). However, I think that in some cases an individual column (or more rarely, an individual neuron) can play a key role in one of these symbol-embodying strange-transients. So, for example, suppose Columns C1, C2, C3 are closely associated with symbol-embodying strange transients T1, T2, T3. Suppose there are highly conductive synaptic bundles going in the directions C1 -- C2 C2 -- C3 Then, Hebbian learning may result in the potentiation of the synaptic bundle going C1 -- C3 Now, we may analyze the relationships between the strange transients T1, T2, T3 using Markov chains, where a high-weight link between T1 and T2, for example, means that P(T2|T1) is large. Then, the above Hebbian learning example will lead to the heuristic inference P(T2 | T1) is large P(T3 | T2) is large |- P(T3 | T1) is large But this is probabilistic term logic deduction (and comes with specific quantitative formulas that I am not giving here). One can make similar analyses for other probabilistic logic rules. Basically, one can ground probabilistic inference on Markov probabilities between strange-transients of the neural network, in Hebbian learning on synaptic bundles between cortical columns. And that is (in very sketchy form, obviously) part of my hypothesis about how the brain may ground symbolic logic in neurodynamics. The subtler part of my hypothesis attempts to explain how higher-order functions and quantified logical relationships may be grounded in neurodynamics. But I don't really want to post that on a list before publishing it formally in a scientific journal, as it's a bigger and also more complex idea. This is not how Novamente works -- Novamente is not a neural net architecture. However, Novamente does include some similar ideas. In Novamente lingo, the strange transients mentioned above are called maps, and the role of the Hebbian learning mentioned above is played in NM by explicit probabilistic term logic. So, according to my view, In the brain: lower-level Hebbian learning on bundles of links btw neuronal clusters, leads to implicit probabilistic inference on strange-transients representing concepts In Novamente: explicit heuristic/probabilistic inference on links btw nodes in NM's hypergraph datastructure, lead to implicit probabilistic inference on strange-transients (called maps) representing concepts So, the Novamente approach seeks to retain the creativity/fluidity-supportive emergence of the brain's approach, while still utilizing a form of probabilistic logic rather than neuron emulations on the lower level. This subtlety causes many people to misunderstand the Novamente architecture, because they only think about the lower level rather than the emergent, map level. In terms of our practical Novamente work we have not done much with the map level yet, but we know this is going to be the crux of the system's AGI capability. -- Ben As I see it, ALL of these choices have their problems. In other words, if the machinery of logical reasoning is actually visible to you in the naked hardware at any of these levels, I reckon that you must then commit to some
[agi] The Missing Piece
Is there anyone out there who has a sense that most of the work being done in AI is still following the same track that has failed for fifty years now? The focus on logic as thought, or neural nets as the bottom-up, brain-imitating solution just isn't getting anywhere? It's the same thing, and it's never getting anywhere. The missing component is thought. What is thought, and how do human beings think? There is no reason that thought cannot be implemented in a sufficiently powerful computing machine -- the problem is how to implement it. Logical deduction or inference is not thought. It is mechanical symbol manipulation that can can be programmed into any scientific pocket calculator. Human intelligence is based on animal intelligence. We can perform logical calculations because we can see the symbols and their relations and move the symbols around in our minds to produce the results, but the intelligence is not the symbol manipulation, but our ability to see the relationships spatialy and decide if the pieces fit correctly throught the process. The world is continuous, spatiotemporal, and non-descrete, and simply is not describable in logical terms. A true AI system has to model the world in the same way -- spatiotemporal sensorimotor maps. Animal intelligence. This is short, and doesn't express my ideas in much detail. But I've been working alone for a long time now, and I think I have to find some people to talk to. I have an AGI project I've been developing, but I can't do it all by myself. If anyone has questions about what alternative ideas I have to the logical paradigm, I can clarify much further, as far as I can. I would just like to maybe make some connections and find some people who aren't stuck in the computational, symbolic mode. Ask some questions, and I'll tell you what I 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/?list_id=303
Re: [agi] The Missing Piece
On Mon, 19 Feb 2007, John Scanlon wrote: ) Is there anyone out there who has a sense that most of the work being ) done in AI is still following the same track that has failed for fifty ) years now? The focus on logic as thought, or neural nets as the ) bottom-up, brain-imitating solution just isn't getting anywhere? It's ) the same thing, and it's never getting anywhere. Yes, they are mostly building robots and trying to pick up blocks or catch balls. Visual perception and motor control for solving this task was first shown in a limited context in the 1960s. You are correct that the bottom up approach is not a theory driven approach. People talk about mystical words, such as Emergence or Complexity, in order to explain how their very simple model of mind can ultimately think like a human. Top-down design of an A.I. requires a theory of what abstract thought processes do. ) The missing component is thought. What is thought, and how do human ) beings think? There is no reason that thought cannot be implemented in ) a sufficiently powerful computing machine -- the problem is how to ) implement it. Right, there are many theories of how to implement an AI. I wouldn't worry too much about trying to define Thought. It has different definitions depending on the different problem solving contexts that it is used. If you focus on making a machine solve problems, then you might see some part of the machine you build will resemble your many uses for the term Thought. ) Logical deduction or inference is not thought. It is mechanical symbol ) manipulation that can can be programmed into any scientific pocket ) calculator. Logical deduction is only one way to think. As you say, there are many other ways to think. Some of these are simple reactive processes, while others are more deliberative and form multistep plans, while still others are reflective and react to problems in actual planning and inference processes. ) Human intelligence is based on animal intelligence. No. Human intelligence has evolved from animal intelligence. Human intelligence is not necessarily a simple subsumption of animal intelligence. ) The world is continuous, spatiotemporal, and non-descrete, and simply is ) not describable in logical terms. A true AI system has to model the ) world in the same way -- spatiotemporal sensorimotor maps. Animal ) intelligence. Logical parts of the world are describable in logical terms. We think in many different ways. Each of these ways uses different representations of the world. We have many specific solutions to specific types of problem solving, but to make a general problem solver we need ways to map these representations from one specific problem solver to another. This allows alternatives to pursue when a specific problem solver gets stuck. This type of robust problem solving requires reasoning by analogy. ) Ask some questions, and I'll tell you what I think. People always have a lot to say, but what we need more of are working algorithms and demonstrations of robust problem solving. - 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/?list_id=303
Re: [agi] The Missing Piece
John Scanlon wrote: Is there anyone out there who has a sense that most of the work being done in AI is still following the same track that has failed for fifty years now? The focus on logic as thought, or neural nets as the bottom-up, brain-imitating solution just isn't getting anywhere? It's the same thing, and it's never getting anywhere. The missing component is thought. What is thought, and how do human beings think? There is no reason that thought cannot be implemented in a sufficiently powerful computing machine -- the problem is how to implement it. Logical deduction or inference is not thought. It is mechanical symbol manipulation that can can be programmed into any scientific pocket calculator. Human intelligence is based on animal intelligence. We can perform logical calculations because we can see the symbols and their relations and move the symbols around in our minds to produce the results, but the intelligence is not the symbol manipulation, but our ability to see the relationships spatialy and decide if the pieces fit correctly throught the process. The world is continuous, spatiotemporal, and non-descrete, and simply is not describable in logical terms. A true AI system has to model the world in the same way -- spatiotemporal sensorimotor maps. Animal intelligence. This is short, and doesn't express my ideas in much detail. But I've been working alone for a long time now, and I think I have to find some people to talk to. I have an AGI project I've been developing, but I can't do it all by myself. If anyone has questions about what alternative ideas I have to the logical paradigm, I can clarify much further, as far as I can. I would just like to maybe make some connections and find some people who aren't stuck in the computational, symbolic mode. Ask some questions, and I'll tell you what I think. John, I have *some* sympathy for what you say, but I am not sure I can buy the commitment to spatiotemporal maps and animal intelligence, because there are many ways to build a mind that do not use symbolic logic, without on the other hand insisting that everything is continuous. You can have discrete symbols, but with internal structure, for example. This is kind of a big, wie open topic, so it might be better for you to write out an essay about what you have in mind when you imagine an alternative approach. Richard Loosemore. - 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/?list_id=303
Re: [agi] The Missing Piece
On 2/19/07, Bo Morgan [EMAIL PROTECTED] wrote: On Mon, 19 Feb 2007, John Scanlon wrote: ) Is there anyone out there who has a sense that most of the work being ) done in AI is still following the same track that has failed for fifty ) years now? The focus on logic as thought, or neural nets as the ) bottom-up, brain-imitating solution just isn't getting anywhere? It's ) the same thing, and it's never getting anywhere. Yes, they are mostly building robots and trying to pick up blocks or catch balls. Visual perception and motor control for solving this task was first shown in a limited context in the 1960s. You are correct that the bottom up approach is not a theory driven approach. People talk about mystical words, such as Emergence or Complexity, in order to explain how their very simple model of mind can ultimately think like a human. Top-down design of an A.I. requires a theory of what abstract thought processes do. ) The missing component is thought. What is thought, and how do human ) beings think? There is no reason that thought cannot be implemented in ) a sufficiently powerful computing machine -- the problem is how to ) implement it. Right, there are many theories of how to implement an AI. I wouldn't worry too much about trying to define Thought. It has different definitions depending on the different problem solving contexts that it is used. If you focus on making a machine solve problems, then you might see some part of the machine you build will resemble your many uses for the term Thought. ) Logical deduction or inference is not thought. It is mechanical symbol ) manipulation that can can be programmed into any scientific pocket ) calculator. Logical deduction is only one way to think. As you say, there are many other ways to think. Some of these are simple reactive processes, while others are more deliberative and form multistep plans, while still others are reflective and react to problems in actual planning and inference processes. ) Human intelligence is based on animal intelligence. No. Human intelligence has evolved from animal intelligence. Human intelligence is not necessarily a simple subsumption of animal intelligence. ) The world is continuous, spatiotemporal, and non-descrete, and simply is ) not describable in logical terms. A true AI system has to model the ) world in the same way -- spatiotemporal sensorimotor maps. Animal ) intelligence. Logical parts of the world are describable in logical terms. We think in many different ways. Each of these ways uses different representations of the world. We have many specific solutions to specific types of problem solving, but to make a general problem solver we need ways to map these representations from one specific problem solver to another. This allows alternatives to pursue when a specific problem solver gets stuck. This type of robust problem solving requires reasoning by analogy. I hope my ignorance does not bother this list too much. Regarding what or what may not be done through logical inference and other expressive enough symbolic approaches; given unlimited resources would it not be possible to implement an UTM with at most a finite overhead which in turn yields that any algorithm running on an UTM could also run on expressive enough symbolic systems, whether they learn or not? I do not argue that it is not inefficient, both for running and implementation speed. It's even so that the logical inference in such a case may be reduced entirely and proven to be more efficiently obviously, than to implement the system direcly on certain systems. I do not think however that such a strict and not well-formulated position is rationally justified since it's not clear (at least not to me) that the logical inference may be efficiently reduced for every algorithm expressed in the logical language. Just rambling and unrelated but perhaps the brain's operations do not even allow for UTMs since they are not so clear and there might not be appropriate transformations and if assume the Turing-Church thesis we might find that there are problems that artificial components may solve that humans cannot even given unlimited resources. Perhaps not very likely since we can simulate the process of an UTM by hand and even the errors may be corrected given enough time. ) Ask some questions, and I'll tell you what I think. People always have a lot to say, but what we need more of are working algorithms and demonstrations of robust problem solving. - 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/?list_id=303 - 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/?list_id=303
Re: [agi] The Missing Piece
It's pretty clear that humans don't run FOPC as a native code, but that we can learn it as a trick. I disagree. I think that Hebbian learning between cortical columns is essentially equivalent to basic probabilistic term logic. Lower-level common-sense inferencing of the Clyde--elephant--gray type falls out of the representations and the associative operations. I think it falls out of the logic of spike timing dependent long term potentiation of bundles of synapses between cortical columns... -- Ben G - 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/?list_id=303
Re: [agi] The Missing Piece
On Monday 19 February 2007 16:08, Ben Goertzel wrote: It's pretty clear that humans don't run FOPC as a native code, but that we can learn it as a trick. I disagree. I think that Hebbian learning between cortical columns is essentially equivalent to basic probabilistic term logic. That's a tantalizing hint (not that I haven't been floating a few of my own :-). I tend to think of my n-D spaces as representing what a column does... CSG is exactly propositional logic if you think of each point as a proposition. It's the mappings between spaces that are the tricky part and give you the equivalent power of predicates, but not in just that form. I haven't looked it, but I'd bet that Hebbian learning is within hollering distance of some of my associative clustering operations, on a conceptual level. I wouldn't try to get NM to represent general knowledge in this way, but, for representing knowledge about the physical environment and things observed and projected therein, having such operations to act on 3D manifolds would be quite valuable True, but I'm envisioning going up to 1-D in some cases. The key problem, vis-a-vis a system that uses symbols as a base representation, is where do the symbols come from? My idea is to generalize operations that do recognition (e.g. of shapes, phonemes) from raw sense data (lots of nerve signals) -- and then to use the same operations all the way up, to form higher-level concepts from patterns of lower-level ones. Once you have symbols, i.e. once you've carved the world into concepts, things get a lot more straightforward. 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/?list_id=303
Re: Mystical Emergence/Complexity [WAS Re: [agi] The Missing Piece]
Bo Morgan wrote: On Mon, 19 Feb 2007, Richard Loosemore wrote: ) Bo Morgan wrote: ) ) On Mon, 19 Feb 2007, John Scanlon wrote: ) ) ) Is there anyone out there who has a sense that most of the work being ) ) done in AI is still following the same track that has failed for ) ) fifty years now? The focus on logic as thought, or neural nets as the ) ) bottom-up, brain-imitating solution just isn't getting anywhere? ) ) It's the same thing, and it's never getting anywhere. ) ) Yes, they are mostly building robots and trying to pick up blocks or catch ) balls. Visual perception and motor control for solving this task was first ) shown in a limited context in the 1960s. You are correct that the bottom up ) approach is not a theory driven approach. People talk about mystical words, ) such as Emergence or Complexity, in order to explain how their very simple ) model of mind can ultimately think like a human. Top-down design of an A.I. ) requires a theory of what abstract thought processes do. ) ) It is interesting that you would say this. ) ) My first reaction was to simply declare that I completely disagree with your ) ...mystical words, such as Emergence or Complexity... comments, but that ) would not have been very constructive. ) ) I am more interested in *why* you would say that. What approaches do you have ) in mind, that are lacking in theory? Who, of all the researchers you had in ) mind, are the ones you most consider to be using those words in a mystical ) way? I think that describing the ways that humans solve problems will help us to understand how they are intelligent. If we have a sufficient description of how humans solve problems then we will have a theory of how humans solve problems. For example, answers to these questions: How do children attach to their parents and not strangers? How do children learn morals and values? How do children learn how to stack blocks? How do children do visual analogy completion problems? How do parents feel anxious when they hear their child crying? Why do our mental processes seem so simple when they are very intricate processes of control, such as making a turn while walking. How do we learn new ways to learn how to think? How do we reflect on our planning mistakes in order to make a better plan next time? We need to describe these processes and view the architecture of human thinking from an implementation point of view. I think that too many people are focusing on simple components that learn to do very simple tasks, such as recognizing handwriting characters or answering questions such as Is there an animal in this picture?. I disagree with an approach that has solved a simple problem and then claims that by massive scaling, massive parallelism, a humanly intelligent thinking process will Emerge. ) More pointedly, would you be able to give a statement of what *they* would ) claim was their most definitive, non-mystical statement of the meaning of ) terms like complexity or emergence, and could you explain why you feel ) that, neverthless, they are saying nothing beyond the vague and mystical? One example of Emergence would be a recurrent neural network that has a given number of stable oscillating states. People use these stable oscillating states instead of using symbols. They invent recurrent neural networks that can transition from one symbol to the next. This is fine work, but we already have symbols and the ability to actually describe human thought in symbolic systems. RNNs have their time and place, but focusing solely on them is a bottom-up approach without a larger theory of mind. Without a larger theory of how humans think these networks will not become humanly intelligent magically. ) I ask this in a genuine spirit of inquiry: I am puzzled as to why people say ) this stuff about complexity, because it has a very, very clear, non-mystical ) meaning. But maybe you are using those words to refer to something different ) than what I mean so I am trying to find out. I'm not saying that complexity is ill-defined. I'm saying that people make a leap such as: Humans are complex systems, which as far as I understand is roughly equivalent to the statement Humans have a lot of degrees of freedom. They use this statement to draw an analogy between a human mind and a neural network with a billion nodes with no description of any organizing structure. What are a few hundred computational elements that a neural network would need to implement? These are the answers to the questions above. That was a surprise: the things that you were referring to when you used the words emergence and complexity are in fact very different from the meanings that a lot of others use, especially when they are making the mystical processes criticism. Your beef is not the same as theirs, by a long way. I work on a complex systems approach to cognition, but from my point of view I am
Re: [agi] The Missing Piece
Ben Goertzel wrote: It's pretty clear that humans don't run FOPC as a native code, but that we can learn it as a trick. I disagree. I think that Hebbian learning between cortical columns is essentially equivalent to basic probabilistic term logic. Lower-level common-sense inferencing of the Clyde--elephant--gray type falls out of the representations and the associative operations. I think it falls out of the logic of spike timing dependent long term potentiation of bundles of synapses between cortical columns... The original suggestion was (IIANM) that humans don't run FOPC as a native code emat the level of symbols and concepts/em (i.e. the concept-stuff that we humans can talk about because we have introspective access at that level of our systems). Now, if you are going to claim that spike-timing-dependent LTP between columns is where some probabilistic term logic is happening ON SYMBOLS, then what you have to do is buy into a story about where symbols are represented and how. I am not clear about whether you are suggesting that the symbols are represented at: (1) the column level, or (2) the neuron level, or (3) the dendritic branch level, or (4) the synapse level, or (perhaps) (5) the spike-train level (i.e. spike trains encode symbol patterns). If you think that the logical machinery is visible, can you say which of these levels is the one where you see it? As I see it, ALL of these choices have their problems. In other words, if the machinery of logical reasoning is actually visible to you in the naked hardware at any of these levels, I reckon that you must then commit to some description of how symbols are implemented, and I think all of them look like bad news. THAT is why, each time the subject is mentioned, I pull a sucking-on-lemons face and start bad-mouthing the neuroscientists. ;-) I don't mind there being some logic-equivalent machinery down there, but I think it would be strictly sub-cognitive, and not relevant to normal human reasoning at all .. and what I find frustrating is that (some of) the people who talk about it seem to think that they only have to find *something* in the neural hardware that can be mapped onto *something* like symbol-manipulation/logical reasoning, and they think they are half way home and dry, without stopping to consider the other implications of the symbols being encoded at that hardware-dependent level. I haven't seen any neuroscientists who talk that way show any indication that they have a clue that there are even problems with it, let alone that they have good answers to those problems. In other words, I don't think I buy it. Richard Loosemore. - 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/?list_id=303
Re: [agi] The Missing Piece
Sorry, I was slow to read. Working on a thought is what makes it maybe one day a realtiy. Nice post. Thanks. Anna:) On 2/19/07, John Scanlon [EMAIL PROTECTED] wrote: Eliezer S. Yudkowsky wrote: John Scanlon wrote: Is there anyone out there who has a sense that most of the work being done in AI is still following the same track that has failed for fifty years now? The focus on logic as thought, or neural nets as the bottom-up, brain-imitating solution just isn't getting anywhere? It's the same thing, and it's never getting anywhere. The missing component is thought. What is thought, and how do human beings think? There is no reason that thought cannot be implemented in a sufficiently powerful computing machine -- the problem is how to implement it. No, that's not it. I know because I once built a machine with thoughts in it and it still didn't work. Do you have any other ideas? Okay, that was a nice, quick dismissive statement. And you're right -- just insert the element of thought, and voila you have intelligence, or in the case of the machine you once built -- nothing. That's not what I mean. I've read some of your stuff, and you know a lot more about computer science and science in general than I may ever know. I don't mean that the missing ingredient is simply the mystical idea of thought. I mean that thought is something different than calculation. Human intelligence is built on animal intelligence -- and what I mean by that is that there was animal intelligence, the same kind of intelligence that can be seen today in apes, before the development of language that was the substrate that allowed the use of language. Language is the manipulation of symbols. When you think of how a non-linguistic proto-human species first started using language, you can imagine creatures associating sounds with images -- oog is the big hairy red ape who's always trying to steal your women. akk is the action of hitting him with a club. The symbol, the sound, is associated with a sensorimotor pattern. The visual pattern is the big hairy red ape you know, and the motor pattern is the sequence of muscle activations that swing the club. In order to use these symbols effectively, you have to have a sensorimotor image or pattern that the symbols are attached to. That's what I'm getting at. That is thought. We already know how to get computers to carry out very complex logical calculations, but it's mechanical, it's not thought, and they can't navigate themselves (with any serious competence) around a playground. Language and logical intelligence is built on visual-spatial modeling. That's why children learn their ABC's by looking at letters drawn on a chalkboard and practicing the muscle movements to draw them on paper. I think that the key to AI is to implement this sensorimotor, spatiotemporal modeling in software. That means data structures that represent the world in three spatial dimensions and one temporal dimension. This modeling can be done. It's done every day in video games. But obviously that's not enough. There is the element of probability -- what usually occurs, what might occur, and how my actions might affect what might occur. Okay -- so what I am focused on is creating data structures that can take sensorimotor patterns and put them into a knowledge-representation system that can remember events, predict events, and predict how motor actions will affect events. And it is all represented in terms of sensorimotor images or maps. I don't have it all figured out right now, but this is what I'm working on. - Original Message - From: Eliezer S. Yudkowsky To: agi@v2.listbox.com Sent: Monday, February 19, 2007 9:12 PM Subject: Re: [agi] The Missing Piece John Scanlon wrote: Is there anyone out there who has a sense that most of the work being done in AI is still following the same track that has failed for fifty years now? The focus on logic as thought, or neural nets as the bottom-up, brain-imitating solution just isn't getting anywhere? It's the same thing, and it's never getting anywhere. The missing component is thought. What is thought, and how do human beings think? There is no reason that thought cannot be implemented in a sufficiently powerful computing machine -- the problem is how to implement it. No, that's not it. I know because I once built a machine with thoughts in it and it still didn't work. Do you have any other ideas? -- Eliezer S. Yudkowsky http://singinst.org/ Research Fellow, Singularity Institute for Artificial Intelligence - 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/?list_id=303 - 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/?list_id=303