Re: [agi] advice-level dev collaboration
Thanks for the responses. Sorry, I picked just a couple of folks. Dealing with the wide audience of the whole AGI list would IMO make things more difficult for me. I may share selected stuff later. Regards, Jiri Jelinek - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=64842712-fa84ee
Re: [agi] Relativistic irrationalism
Stefan, Though I agree with most of your analysis on inter-agent relationship, I don't share your conception of rationality. To me, rationality itself is relativistic, that is, what behavior/action is rational is always judged according to the assumptions and postulations on a system's goal, knowledge, resources, etc. There is no single rationality that can be used in all situations. Similar ideas have been argued by I.J. Good, H.A. Simon, and some others. In the context of AGI, AIXI is an important model of rationality, but not the only one. At least there are NARS and OSCAR, which are based on different assumptions about the system and its environment. Being impractical is not the only problem of AIXI. As soon as one of its assumptions (infinite resources is only one of them) is dropped, its conclusions become inapplicable. Some people think in theory we should accept unrealistic assumptions, like infinite resources, since they lead to rigorous models; then, in implementation, the realistic restrictions (on resources etc.) can be introduced, which lead to approximations of the idealized model. What they fail to see is that when a new restriction is added, it may change the problem to the extent that the ideal theory becomes mostly irrelevant. To me, it is much better to start with more realistic assumptions in the first place, even though it will make the problem harder to solve. Pei On Nov 13, 2007 10:40 PM, Stefan Pernar [EMAIL PROTECTED] wrote: Would be great if people could poke the following with their metaphorical sticks: Imagine two agents A(i) each one with a utility function F(i), capability level C(i) and no knowledge as to the other agents F and C values. Both agents are given equal resources and are tasked with devising the most efficient and effective way to maximize their respective utility with said resources. Scenario 1: Both agents have fairly similar utility functions F(1) = F(2), level of knowledge, cognitive complexity, experience - in short capability C(1) = C(2) - and a high level of mutual trust T(1-2) = T(2-1) = 1. They will quickly agree on the way forward, pool their resources and execute their joint plan. Rather boring. Scenario 2: Again we assume F(1) = F(2), however C(1) C(2) - again T(1-2) = T(2-1) = 1. The more capable agent will devise a plan, the less capable agent will provide its resources and execute the plan trusted by C(2). A bit more interesting. Scenario 3: F(1) = F(2), C(1) C(2) but this time T(1-2) = 1 and T(2-1) = 0.5 meaning the less powerful agent assumes with a probability of 50% that A(1) is in fact a self serving optimizer who's difference in plan will turn out to be decremental to A(2) while A(1) is certain that this is all just one big misunderstanding. The optimal plan devised under scenario 2 will now face opposition by A(2) although it would be in A(2)'s best interest to actually support it with its resources to maximize (F2) while A(1) will see A(2)'s objection as being detrimental to maximizing their shared utility function. Fairly interesting: based on lack of trust and differences in capability each agent perceives the other agent's plan as being irrational from their respective points of view. Under scenario 3, both agents now have a variety of strategies at their disposal: deny pooling of part or all of ones resources = If we do not do it my way you can do it alone. use resources to sabotage the other agent's plan = I must stop him with these crazy ideas! deceive the other agent in order to skew how the other agent is deploying strategies 1 and 2 spend resources to explain the plan to the other agent = Ok - let's help him see the light spend resources on self improvement to understand the other agent's plan better = Let's have a closer look, the plan might not be so bad after all strike a compromise to ensure a higher level of pooled resources = If we don't compromise we both loose out Number 1 is a given under scenario 3. Number 2 is risky, particularly as it would cause a further reduction in trust on both sides if this strategy gets deployed assuming the other party would find out similarly with number 3. Number 4 seems like the way to go but may not always work particularly with large differences in C(i) among the agents. Number 5 is a likely strategy with a fairly high level of trust. Most likely however is strategy 6. Striking a compromise is trust building in repeated encounters and thus promises less objection and thus higher total payoff the next times around. Assuming the existence of an arguably optimal path leading to a maximally possible satisfaction of a given utility function anything else would be irrational. Actually such a maximally intelligent algorithm exists in the form of Hutter's universal algorithmic agent AIXI. The only problem being however that the execution of said algorithm requires infinite resources and is thus rather unpractical as every
Re: [agi] What best evidence for fast AI?
I think that there are two basic directions to better the Novamente architecture: the one Mark talks about more integration of MOSES with PLN and RL theory On 11/13/07, Edward W. Porter [EMAIL PROTECTED] wrote: Response to Mark Waser Mon 11/12/2007 2:42 PM post. MARK Remember that the brain is *massively* parallel. Novamente and any other linear (or minorly-parallel) system is *not* going to work in the same fashion as the brain. Novamente can be parallelized to some degree but *not* to anywhere near the same degree as the brain. I love your speculation and agree with it -- but it doesn't match near-term reality. We aren't going to have brain-equivalent parallelism anytime in the near future. ED I think in five to ten years there could be computers capable of providing every bit as much parallelism as the brain at prices that will allow thousands or hundreds of thousands of them to be sold. But it is not going to happen overnight. Until then the lack of brain level hardware is going to limit AGI. But there are still a lot of high value system that could be built on say $100K to $10M of hardware. You claim we really need experience with computing and controlling activation over large atom tables. I would argue that obtaining such experience should be a top priority for government funders. MARK The node/link architecture is very generic and can be used for virtually anything. There is no rational way to attack it. It is, I believe, going to be the foundation for any system since any system can easily be translated into it. Attacking the node/link architecture is like attacking assembly language or machine code. Now -- are you going to write your AGI in assembly language? If you're still at the level of arguing node/link, we're not communicating well. ED nodes and links are what patterns are made of, and each static pattern can have an identifying node associated with it as well as the nodes and links representing its sub-patterns, elements, the compositions of which it is part, it associations, etc. The system automatically organize patterns into a gen/comp hierarchy. So, I am not just dealing at a node and link level, but they are the basic building blocks. MARK ... I *AM* saying that the necessity of using probabilistic reasoning for day-to-day decision-making is vastly over-rated and has been a horrendous side-road for many/most projects because they are attempting to do it in situations where it is NOT appropriate. The increased, almost ubiquitous adaptation of probabilistic methods is the herd mentality in action (not to mention the fact that it is directly orthogonal to work thirty years older). Most of the time, most projects are using probabilistic methods to calculate a tenth place decimal of a truth value when their data isn't even sufficient for one. If you've got a heavy-duty discovery system, probabilistic methods are ideal. If you're trying to derive probabilities from a small number of English statements (like this raven is white and most ravens are black), you're seriously on the wrong track. If you go on and on about how humans don't understand Bayesian reasoning, you're both correct and clueless in not recognizing that your very statement points out how little Bayesian reasoning has to do with most general intelligence. Note, however, that I *do* believe that probabilistic methods *are* going to be critically important for activation for attention, etc. ED I agree that many approaches accord too much importance to the numerical accuracy and Bayesian purity of their approach, and not enough importance on the justification for the Bayesian formulations they use. I know of one case where I suggested using information that would almost certainly have improved a perception process and the suggestion was refused because it would not fit within the system's probabilistic framework. At an AAAI conference in 1997 I talked to a programmer for a big defense contractor who said he as a fan of fuzzy logic system; that they were so much more simple to get up an running because you didn't have to worry about probabilistic purity. He said his group that used fuzzy logic was getting things out the door that worked faster than the more probability limited competition. So obviously there is something to say for not letting probabilistic purity get in the way of more reasonable approaches. But I still think probabilities are darn important. Even your this raven is white and most ravens are black example involves notions of probability. We attribute probabilities to such statements based on experience with the source of such statements or similar sources of information, and the concept most is a probabilistic one. The reason we humans are so good at reasoning from small data is based on our ability to estimate rough probabilities from similar or generic patterns. MARK The
RE: [agi] What best evidence for fast AI?
Lukasz, Which of the multiple issues that Mark listed is one of the two basic directions you were referring to. Ed Porter -Original Message- From: Lukasz Stafiniak [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 14, 2007 9:15 AM To: agi@v2.listbox.com Subject: Re: [agi] What best evidence for fast AI? I think that there are two basic directions to better the Novamente architecture: the one Mark talks about more integration of MOSES with PLN and RL theory On 11/13/07, Edward W. Porter [EMAIL PROTECTED] wrote: Response to Mark Waser Mon 11/12/2007 2:42 PM post. MARK Remember that the brain is *massively* parallel. Novamente MARK and any other linear (or minorly-parallel) system is *not* going to work in the same fashion as the brain. Novamente can be parallelized to some degree but *not* to anywhere near the same degree as the brain. I love your speculation and agree with it -- but it doesn't match near-term reality. We aren't going to have brain-equivalent parallelism anytime in the near future. ED I think in five to ten years there could be computers capable ED of providing every bit as much parallelism as the brain at prices that will allow thousands or hundreds of thousands of them to be sold. But it is not going to happen overnight. Until then the lack of brain level hardware is going to limit AGI. But there are still a lot of high value system that could be built on say $100K to $10M of hardware. You claim we really need experience with computing and controlling activation over large atom tables. I would argue that obtaining such experience should be a top priority for government funders. MARK The node/link architecture is very generic and can be used MARK for virtually anything. There is no rational way to attack it. It is, I believe, going to be the foundation for any system since any system can easily be translated into it. Attacking the node/link architecture is like attacking assembly language or machine code. Now -- are you going to write your AGI in assembly language? If you're still at the level of arguing node/link, we're not communicating well. ED nodes and links are what patterns are made of, and each static pattern can have an identifying node associated with it as well as the nodes and links representing its sub-patterns, elements, the compositions of which it is part, it associations, etc. The system automatically organize patterns into a gen/comp hierarchy. So, I am not just dealing at a node and link level, but they are the basic building blocks. MARK ... I *AM* saying that the necessity of using probabilistic reasoning for day-to-day decision-making is vastly over-rated and has been a horrendous side-road for many/most projects because they are attempting to do it in situations where it is NOT appropriate. The increased, almost ubiquitous adaptation of probabilistic methods is the herd mentality in action (not to mention the fact that it is directly orthogonal to work thirty years older). Most of the time, most projects are using probabilistic methods to calculate a tenth place decimal of a truth value when their data isn't even sufficient for one. If you've got a heavy-duty discovery system, probabilistic methods are ideal. If you're trying to derive probabilities from a small number of English statements (like this raven is white and most ravens are black), you're seriously on the wrong track. If you go on and on about how humans don't understand Bayesian reasoning, you're both correct and clueless in not recognizing that your very statement points out how little Bayesian reasoning has to do with most general intelligence. Note, however, that I *do* believe that probabilistic methods *are* going to be critically important for activation for attention, etc. ED I agree that many approaches accord too much importance to the numerical accuracy and Bayesian purity of their approach, and not enough importance on the justification for the Bayesian formulations they use. I know of one case where I suggested using information that would almost certainly have improved a perception process and the suggestion was refused because it would not fit within the system's probabilistic framework. At an AAAI conference in 1997 I talked to a programmer for a big defense contractor who said he as a fan of fuzzy logic system; that they were so much more simple to get up an running because you didn't have to worry about probabilistic purity. He said his group that used fuzzy logic was getting things out the door that worked faster than the more probability limited competition. So obviously there is something to say for not letting probabilistic purity get in the way of more reasonable approaches. But I still think probabilities are darn important. Even your this raven is white and most ravens are black example involves notions of probability. We attribute
Re: [agi] Relativistic irrationalism
Pei, many thanks for your comments. Good input on rationality and AIXI. Kind regards, Stefan On Nov 14, 2007 10:13 PM, Pei Wang [EMAIL PROTECTED] wrote: Stefan, Though I agree with most of your analysis on inter-agent relationship, I don't share your conception of rationality. To me, rationality itself is relativistic, that is, what behavior/action is rational is always judged according to the assumptions and postulations on a system's goal, knowledge, resources, etc. There is no single rationality that can be used in all situations. Similar ideas have been argued by I.J. Good, H.A. Simon, and some others. In the context of AGI, AIXI is an important model of rationality, but not the only one. At least there are NARS and OSCAR, which are based on different assumptions about the system and its environment. Being impractical is not the only problem of AIXI. As soon as one of its assumptions (infinite resources is only one of them) is dropped, its conclusions become inapplicable. Some people think in theory we should accept unrealistic assumptions, like infinite resources, since they lead to rigorous models; then, in implementation, the realistic restrictions (on resources etc.) can be introduced, which lead to approximations of the idealized model. What they fail to see is that when a new restriction is added, it may change the problem to the extent that the ideal theory becomes mostly irrelevant. To me, it is much better to start with more realistic assumptions in the first place, even though it will make the problem harder to solve. Pei On Nov 13, 2007 10:40 PM, Stefan Pernar [EMAIL PROTECTED] wrote: Would be great if people could poke the following with their metaphorical sticks: Imagine two agents A(i) each one with a utility function F(i), capability level C(i) and no knowledge as to the other agents F and C values. Both agents are given equal resources and are tasked with devising the most efficient and effective way to maximize their respective utility with said resources. Scenario 1: Both agents have fairly similar utility functions F(1) = F(2), level of knowledge, cognitive complexity, experience - in short capability C(1) = C(2) - and a high level of mutual trust T(1-2) = T(2-1) = 1. They will quickly agree on the way forward, pool their resources and execute their joint plan. Rather boring. Scenario 2: Again we assume F(1) = F(2), however C(1) C(2) - again T(1-2) = T(2-1) = 1. The more capable agent will devise a plan, the less capable agent will provide its resources and execute the plan trusted by C(2). A bit more interesting. Scenario 3: F(1) = F(2), C(1) C(2) but this time T(1-2) = 1 and T(2-1) = 0.5 meaning the less powerful agent assumes with a probability of 50% that A(1) is in fact a self serving optimizer who's difference in plan will turn out to be decremental to A(2) while A(1) is certain that this is all just one big misunderstanding. The optimal plan devised under scenario 2 will now face opposition by A(2) although it would be in A(2)'s best interest to actually support it with its resources to maximize (F2) while A(1) will see A(2)'s objection as being detrimental to maximizing their shared utility function. Fairly interesting: based on lack of trust and differences in capability each agent perceives the other agent's plan as being irrational from their respective points of view. Under scenario 3, both agents now have a variety of strategies at their disposal: deny pooling of part or all of ones resources = If we do not do it my way you can do it alone. use resources to sabotage the other agent's plan = I must stop him with these crazy ideas! deceive the other agent in order to skew how the other agent is deploying strategies 1 and 2 spend resources to explain the plan to the other agent = Ok - let's help him see the light spend resources on self improvement to understand the other agent's plan better = Let's have a closer look, the plan might not be so bad after all strike a compromise to ensure a higher level of pooled resources = If we don't compromise we both loose out Number 1 is a given under scenario 3. Number 2 is risky, particularly as it would cause a further reduction in trust on both sides if this strategy gets deployed assuming the other party would find out similarly with number 3. Number 4 seems like the way to go but may not always work particularly with large differences in C(i) among the agents. Number 5 is a likely strategy with a fairly high level of trust. Most likely however is strategy 6. Striking a compromise is trust building in repeated encounters and thus promises less objection and thus higher total payoff the next times around. Assuming the existence of an arguably optimal path leading to a maximally possible satisfaction of a given utility function anything else would be
Re: [agi] What best evidence for fast AI?
On Nov 14, 2007 3:48 PM, Edward W. Porter [EMAIL PROTECTED] wrote: Lukasz, Which of the multiple issues that Mark listed is one of the two basic directions you were referring to. Ed Porter (First of all, I'm sorry for attaching my general remark as a reply: I was writing from a cell-phone which limited navigation.) I think, that it would be a more fleshed-out knowledge representation (but without limiting the representation-building flexibility of Novamente). -Original Message- From: Lukasz Stafiniak [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 14, 2007 9:15 AM To: agi@v2.listbox.com Subject: Re: [agi] What best evidence for fast AI? I think that there are two basic directions to better the Novamente architecture: the one Mark talks about more integration of MOSES with PLN and RL theory - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=64970556-f74c23
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
Linas Vepstas wrote: On Tue, Nov 13, 2007 at 12:34:51PM -0500, Richard Loosemore wrote: Suppose that in some significant part of Novamente there is a representation system that uses probability or likelihood numbers to encode the strength of facts, as in [I like cats](p=0.75). The (p=0.75) is supposed to express the idea that the statement [I like cats] is in some sense 75% true. Either way, we have a problem: a fact like [I like cats](p=0.75) is ungrounded because we have to interpret it. Does it mean that I like cats 75% of the time? That I like 75% of all cats? 75% of each cat? Are the cats that I like always the same ones, or is the chance of an individual cat being liked by me something that changes? Does it mean that I like all cats, but only 75% as much as I like my human family, which I like(p=1.0)? And so on and so on. Eh? You are standing at the proverbial office water coooler, and Aneesh says Wen likes cats. On your drive home, you mind races .. does this mean that Wen is a cat fancier? You were planning on taking Wen out on a date, and this tidbit of information could be useful ... when you try to build the entire grounding mechanism(s) you are forced to become explicit about what these numbers mean, during the process of building a grounding system that you can trust to be doing its job: you cannot create a mechanism that you *know* is constructing sensible p numbers and facts during all of its development *unless* you finally bite the bullet and say what the p numbers really mean, in fully cashed out terms. But has a human, asking Wen out on a date, I don't really know what Wen likes cats ever really meant. It neither prevents me from talking to Wen, or from telling my best buddy that ...well, I know, for instance, that she likes cats... Lack of grounding is what makes humour funny, you can do a whole Pygmalion / Seinfeld episode on she likes cats. No: the real concept of lack of grounding is nothing so simple as the way you are using the word grounding. Lack of grounding makes an AGI fall flat on its face and not work. I can't summarize the grounding literature in one post. (Though, heck, I have actually tried to do that in the past: didn't do any good). 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/?member_id=8660244id_secret=64980585-67cbc9
Re: Introducing Autobliss 1.0 (was RE: [agi] Nirvana? Manyana? Never!)
Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: Matt Mahoney wrote: --- Jiri Jelinek [EMAIL PROTECTED] wrote: On Nov 11, 2007 5:39 PM, Matt Mahoney [EMAIL PROTECTED] wrote: We just need to control AGIs goal system. You can only control the goal system of the first iteration. ..and you can add rules for it's creations (e.g. stick with the same goals/rules unless authorized otherwise) You can program the first AGI to program the second AGI to be friendly. You can program the first AGI to program the second AGI to program the third AGI to be friendly. But eventually you will get it wrong, and if not you, then somebody else, and evolutionary pressure will take over. This statement has been challenged many times. It is based on assumptions that are, at the very least, extremely questionable, and according to some analyses, extremely unlikely. I guess it will continue to be challenged until we can do an experiment to prove who is right. Perhaps you should challenge SIAI, since they seem to think that friendliness is still a hard problem. I have done so, as many people on this list will remember. The response was deeply irrational. 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/?member_id=8660244id_secret=64985895-75bf5b
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
Hi, No: the real concept of lack of grounding is nothing so simple as the way you are using the word grounding. Lack of grounding makes an AGI fall flat on its face and not work. I can't summarize the grounding literature in one post. (Though, heck, I have actually tried to do that in the past: didn't do any good). FYI, I have read the symbol-grounding literature (or a lot of it), and generally found it disappointingly lacking in useful content... though I do agree with the basic point that non-linguistic grounding is extremely helpful for effective manipulation of linguistic entities... -- 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/?member_id=8660244id_secret=64981284-09925d
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
Benjamin Goertzel wrote: On Nov 13, 2007 2:37 PM, Richard Loosemore [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] wrote: Ben, Unfortunately what you say below is tangential to my point, which is what happens when you reach the stage where you cannot allow any more vagueness or subjective interpretation of the qualifiers, because you have to force the system to do its own grounding, and hence its own interpretation. I don't see why you talk about forcing the system to do its own grounding -- the probabilities in the system are grounded in the first place, as they are calculated based on experience. The system observes, records what it sees, abstracts from it, and chooses actions that it guess will fulfill its goals. Its goals are ultimately grounded in in-built feeling-evaluation routines, measuring stuff like amount of novelty observed, amount of food in system etc. So, the system sees and then acts ... and the concepts it forms and uses are created/used based on their utility in deriving appropriate actions. There is no symbol-grounding problem except in the minds of people who are trying to interpret what the system does, and get confused. Any symbol used within the system, and any probability calculated by the system, are directly grounded in the system's experience. There is nothing vague about an observation like Bob_Yifu was observed at time-stamp 599933322, or a fact Command 'wiggle ear' was sent at time-stamp 54. These perceptions and actions are the root of the probabilities the system calculated, and need no further grounding. What you gave below was a sketch of some more elaborate 'qualifier' mechanisms. But I described the process of generating more and more elaborate qualifier mechanisms in the body of the essay, and said why this process was of no help in resolving the issue. So, if a system can achieve its goals based on choosing procedures that it thinks are likely to achieve its goals, based on the knowledge it gathered via its perceived experience -- why do you think it has a problem? I don't really understand your point, I guess. I thought I did -- I thought your point was that precisely specifying the nature of a conditional probability is a rats-nest of complexity. And my response was basically that in Novamente we don't need to do that, because we define conditional probabilities based on the system's own knowledge-base, i.e. Inheritance A B .8 means If A and B were reasoned about a lot, then A would (as measred by an weighted average) have 80% of the relationships that B does But apparently you were making some other point, which I did not grok, sorry... Anyway, though, Novamente does NOT require logical relations of escalating precision and complexity to carry out reasoning, which is one thing you seemed to be assuming in your post. You are, in essence, using one of the trivial versions of what symbol grounding is all about. The complaint is not your symbols are not connected to experience. Everyone and their mother has an AI system that could be connected to real world input. The simple act of connecting to the real world is NOT the core problem. If you have an AGI system in which the system itself is allowed to do all the work of building AND interpreting all of its symbols, I don't have any issues with it. Where I do have an issue is with a system which is supposed to be doing the above experiential pickup, and where the symbols are ALSO supposed to be interpretable by human programmers who are looking at things like probability values attached to facts. When a programmer looks at a situation like ContextLink .7,.8 home InheritanceLink Bob_Yifu friend ... and then follows this with a comment like: which suggests that Bob is less friendly at home than in general. ... they have interpreted the meaning of that statement using their human knowledge. So here I am, looking at this situation, and I see: AGI system intepretation (implicit in system use of it) Human programmer intepretation and I ask myself which one of these is the real interpretation? It matters, because they do not necessarily match up. The human programmer's intepretation has a massive impact on the system because all the inference and other mechanisms are built around the assumption that the probabilities mean a certain set of things. You manipulate those p values, and your manipulations are based on assumptions about what they mean. But if the system is allowed to pick up its own knowledge from the environment, the implicit meaning of those p values will not necessarily match the human interpretation. As I say, the meaning is then implicit in the way the system *uses* those p values (and other stuff). It is a nontrivial question to ask whether the implicit system interpretation does indeed match the human intepretation built into the inference
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
Benjamin Goertzel wrote: Hi, No: the real concept of lack of grounding is nothing so simple as the way you are using the word grounding. Lack of grounding makes an AGI fall flat on its face and not work. I can't summarize the grounding literature in one post. (Though, heck, I have actually tried to do that in the past: didn't do any good). FYI, I have read the symbol-grounding literature (or a lot of it), and generally found it disappointingly lacking in useful content... though I do agree with the basic point that non-linguistic grounding is extremely helpful for effective manipulation of linguistic entities... Ben, As you will recall, Harnad himself got frustrated with the many people who took the term symbol grounding and trivialized or distorted it in various ways. One of the reasons the grounding literature is such a waste of time (and you are right: it is) is that so many people talked so much nonsense about it. As far as I am concerned, your use of it is one of those trivial senses that Harnad complained of. (Essentially, if the system uses world input IN ANY WAY during the building of its symbols, then the system is grounded). The effort I put into that essay yesterday will have been completely wasted if your plan is to stick to that interpretation and not discuss the deeper issue that I raised. I really have no energy for pursuing yet another discussion about symbol grounding. Sorry: don't mean to blow you off, but you and I both have better things to do, and I foresee a big waste of time ahead if we pursue it. So let's just drop 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/?member_id=8660244id_secret=64998305-6bdb18
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
Richard,
So here I am, looking at this situation, and I see:
AGI system intepretation (implicit in system use of it)
Human programmer intepretation
and I ask myself which one of these is the real interpretation?
It matters, because they do not necessarily match up.
That is true, but in some cases they may approximate each other well..
In others, not...
This happens to be a pretty simple case, so the odds of a good
approximation seem high.
The human
programmer's intepretation has a massive impact on the system because
all the inference and other mechanisms are built around the assumption
that the probabilities mean a certain set of things. You manipulate
those p values, and your manipulations are based on assumptions about
what they mean.
Well, the PLN inference engine's treatment of
ContextLink
home
InheritanceLink Bob_Yifu friend
is in no way tied to whether the system's implicit interpretation of the
ideas of home or friend are humanly natural, or humanly comprehensible.
The same inference rules will be applied to cases like
ContextLink
Node_66655
InheritanceLink Bob_Yifu Node_544
where the concepts involved have no humanly-comprehensible label.
It is true that the interpretation of ContextLink and InheritanceLink are
fixed
by the wiring of the system, in a general way (but what kinds of properties
are referred to by them may vary in a way dynamically determined by the
system).
In order to completely ground the system, you need to let the system
build its own symbols, yes, but that is only half the story: if you
still have a large component of the system that follows a
programmer-imposed interpretation of things like probability values
attached to facts, you have TWO sets of symbol-using mechanisms going
on, and the system is not properly grounded (it is using both grounded
and ungrounded symbols within one mechanism).
I don't think the system needs to learn its own probabilistic reasoning
rules
in order to be an AGI. This, to me, is too much like requiring that a brain
needs
to learn its own methods for modulating the conductances of the bundles of
synapses linking between the neurons in cell assembly A and cell assembly B.
I don't see a problem with the AGI system having hard-wired probabilistic
inference rules, and hard-wired interpretations of probabilistic link
types. But
the interpretation of any **particular** probabilistic relationship inside
the system, is relative
to the concepts and the empirical and conceptual relationships that the
system
has learned.
You may think that the brain learns its own uncertain inference rules based
on a
lower-level infrastructure that operates in terms entirely unconnected from
ideas
like uncertainty and inference. I think this is wrong. I think the brain's
uncertain
inference rules are the result, on the cell assembly level, of Hebbian
learning and
related effects on the neuron/synapse level. So I think the brain's basic
uncertain
inference rules are wired-in, just as Novamente's are, though of course
using
a radically different infrastructure.
Ultimately an AGI system needs to learn its own reasoning rules and
radically
modify and improve itself, if it's going to become strongly superhuman! But
that is
not where we need to start...
-- 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/?member_id=8660244id_secret=64998317-8c4281
Re: [agi] What best evidence for fast AI?
Bryan Bishop wrote: On Tuesday 13 November 2007 09:11, Richard Loosemore wrote: This is the whole brain emulation approach, I guess (my previous comments were about evolution of brains rather than neural level duplication). Ah, you are right. But this too is an interesting topic. I think that the order of magnitudes for whole brain emulation, connectome, and similar evolutionary methods, are roughly the same, but I haven't done any calculations. It seems quite possible that what we need is a detailed map of every synapse, exact layout of dendritic tree structures, detailed knowledge of the dynamics of these things (they change rapidly) AND wiring between every single neuron. Hm. It would seem that we could have some groups focusing on neurons, another on types of neurons, another on dendritic tree structures, some more on the abstractions of dendritic trees, etc. in an up-*and*-down propagation hierarchy so that the abstract processes of the brain are studied just as well as the in-betweens of brain architecture. I was really thinking of the data collection problem: we cannot take one brain and get full information about all those things, down to a sufficient level of detail. I do not see such a technology even over the horizon (short of full-blow nanotechnology) that can deliver that. We can get different information from different individual brains (all of them dead), but combining that would not necessarily be meaningful: all brains are different. I think that if they did the whole project at that level of detail it would amount to a possibly interesting hint at some of the wiring, of peripheral interest to people doing work at the cognitive system level. But that is all. You see no more possible value of such a project? Well, I think that it will have more value one day, but at such a late stage in the history of cognitive system building that it will essentially just be a mopping up operation. In other words, we will have to do so much work at the cognitive level to be able to make sense of the wiring diagrams, that by that stage we will be able to generate our own systems. 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/?member_id=8660244id_secret=65002389-10cd4a
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
RL:In order to completely ground the system, you need to let the system build its own symbols V. much agree with your whole argument. But - I may well have missed some vital posts - I have yet to get the slightest inkling of how you yourself propose to do this. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=65013351-96e8f0
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
On Nov 14, 2007 1:36 PM, Mike Tintner [EMAIL PROTECTED] wrote: RL:In order to completely ground the system, you need to let the system build its own symbols Correct. Novamente is designed to be able to build its own symbols. what is built-in, are mechanisms for building symbols, and for probabilistically interrelating symbols once created... ben g V. much agree with your whole argument. But - I may well have missed some vital posts - I have yet to get the slightest inkling of how you yourself propose to do this. - 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/?; - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=65100803-21ddd3
Re: [agi] What best evidence for fast AI?
On Wednesday 14 November 2007 11:55, Richard Loosemore wrote: I was really thinking of the data collection problem: we cannot take one brain and get full information about all those things, down to a sufficient level of detail. I do not see such a technology even over the horizon (short of full-blow nanotechnology) that can deliver that. We can get different information from different individual brains (all of them dead), but combining that would not necessarily be meaningful: all brains are different. Re: all brains are different. What about the possibilities of cloning mice and then proceeding to raise them in Skinner boxes with the exact same environmental conditions, the same stimulation routines, etc. ? Ideally this will give us a baseline mouse that is not only genetically similar, but also behaviorally similar to some degree. This would undoubtedly be helpful in this quest. - Bryan - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=65191157-9f3b24
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
On Wednesday 14 November 2007 11:28, Richard Loosemore wrote: The complaint is not your symbols are not connected to experience. Everyone and their mother has an AI system that could be connected to real world input. The simple act of connecting to the real world is NOT the core problem. Are we sure? How much of the real world are we able to get into our AGI models anyway? Bandwidth is limited, much more limited than in humans and other animals. In fact, it might be the equivalent to worm tech. To do the calculations would I just have to check out how many neurons are in a worm, how many sensory neurons, and rough information theoretic estimations as to the minimum and maximums as to amounts of information processing that the worm's sensorium could be doing? - Bryan - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=65191610-b12544
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
Bryan Bishop wrote: On Wednesday 14 November 2007 11:28, Richard Loosemore wrote: The complaint is not your symbols are not connected to experience. Everyone and their mother has an AI system that could be connected to real world input. The simple act of connecting to the real world is NOT the core problem. Are we sure? How much of the real world are we able to get into our AGI models anyway? Bandwidth is limited, much more limited than in humans and other animals. In fact, it might be the equivalent to worm tech. To do the calculations would I just have to check out how many neurons are in a worm, how many sensory neurons, and rough information theoretic estimations as to the minimum and maximums as to amounts of information processing that the worm's sensorium could be doing? I'm not quite sure where this is at . but the context of this particular discussion is the notion of 'symbol grounding' raised by Steven Harnad. I am essentially talking about how to solve the problem he described, and what exactly the problem was. Hence a lot of background behind this one, which if you don't know it might make it confusing. 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/?member_id=8660244id_secret=65202116-6cf6d0
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
On Nov 14, 2007 11:58 PM, Bryan Bishop [EMAIL PROTECTED] wrote: Are we sure? How much of the real world are we able to get into our AGI models anyway? Bandwidth is limited, much more limited than in humans and other animals. In fact, it might be the equivalent to worm tech. To do the calculations would I just have to check out how many neurons are in a worm, how many sensory neurons, and rough information theoretic estimations as to the minimum and maximums as to amounts of information processing that the worm's sensorium could be doing? Pretty much. Let's take as our reference computer system a bog standard video camera connected to a high-end PC, which can do something (video compression, object recognition or whatever) with the input in real time. On the worm side, consider the model organism Caenorhabditis elegans, which has a few hundred neurons. It turns out that the computer has much more bandwidth. Then again, while intelligence unlike bandwidth isn't a scalar quantity even to a first approximation, to the extent they are comparable our best computer systems do seem to be considerably smarter than C. elegans. If we move up to something like a mouse, then the mouse has intelligence we can't replicate, and also has much more bandwidth than the computer system. Insects are somewhere in between, enough so that the comparison (both bandwidth and intelligence) doesn't produce an obvious answer; it's therefore considered not unreasonable to say present-day computers are in the ballpark of insect-smart. Of course that doesn't mean if we took today's software and connected it to mouse-bandwidth hardware it would become mouse-smart, but hopefully it means when we have that hardware we'll be able to use it to develop software that matches some of the things mice can do. (And it's still my opinion that by accepting - embracing - slowness on existing hardware we can work on the software at the same time as the hardware guys are working on their end, parallel rather than serial development.) - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=65207531-031731
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
Sounds a little confusing. Sounds like you plan to evolve a system through testing thousands of candidate mechanisms. So one way or another you too are taking a view - even if it's an evolutionary, I'm not taking a view view - on, and making a lot of asssumptions about -how systems evolve -the known architecture of human cognition. about which science has extremely patchy and confused knowledge. I don't see how any system-builder can avoid taking a view of some kind on such matters, yet you seem to be criticising Ben for so doing. I was hoping that you also had some view on how a system 's symbols should be grounded, especially since you mention Harnad, who does make vague gestures towards the brain's levels of grounding. But you don't indicate any such view. Sounds like you too, pace MW, are hoping for a number of miracles - IOW creative ideas - to emerge, and make your system work. Anyway, you have to give Ben credit for putting a lot of his stuff principles out there on the line. I think anyone who wants to mount a full-scale assault on him ( why not?) should be prepared to reciprocate. - RL: Mike Tintner wrote: RL:In order to completely ground the system, you need to let the system build its own symbols V. much agree with your whole argument. But - I may well have missed some vital posts - I have yet to get the slightest inkling of how you yourself propose to do this. Well, for the purposes of the present discussion I do not need to say how, only to say that there is a difference between two different research strategies for finding out what the mechanism is that does this. One strategy (the one that I claim has serious problems) is where you try to have your cake and eat it too: let the system build its own symbols, with attached parameters that 'mean' whatever they end up meaning after the symbols have been built, BUT then at the same time insist that some of the parameters really do 'mean' things like probabilities or likelihood or confidence values. If the programmer does anything at all to include mechanisms that rely on these meanings (these interpretations of what the parameters signify) then the programmer has second-guessed what the system itself was going to use those things for, and you have a conflict between the two. My strategy is to keep my hands off, not do anything to strictly interpret those parameters, and experimentally observe the properties of systems that seem loosely consistent with the known architecture of human cognition. I have a parameter, for instance, that seems to be a happiness or consistency parameter attached to a knowledge-atom. But beyond roughly characterising it as such, I do not insert any mechanisms that (implicitly or explicitly) lock the system into such an intepretation. Instead, I have a wide variety of different candidate mechanisms that use that parameter, and I look at the overall properties of systems that use these different candidate mechanisms. I let the system use the parameter according to the dictates of whatever mechanism is in place, but then I just explore the consequences (the high level behavior of the system). In this way I do not get a conflict between what I think the parameter 'ought' to mean and what the system is implicitly taking it to 'mean' by its use of the parameter. I could start talking about all the different candidate mechanisms, but there are thousands of them (at least thousands of candidates that I go so far as to test: they are generated in a semi-automatic way, so there are an unlimited number of potential candidates). 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/?; -- No virus found in this incoming message. Checked by AVG Free Edition. Version: 7.5.503 / Virus Database: 269.15.30/1125 - Release Date: 11/11/2007 9:50 PM - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=65232546-91c089
Re: Essay - example of how the CSP bites [WAS Re: [agi] What best evidence for fast AI?]
Mike Tintner wrote: RL:In order to completely ground the system, you need to let the system build its own symbols V. much agree with your whole argument. But - I may well have missed some vital posts - I have yet to get the slightest inkling of how you yourself propose to do this. Well, for the purposes of the present discussion I do not need to say how, only to say that there is a difference between two different research strategies for finding out what the mechanism is that does this. One strategy (the one that I claim has serious problems) is where you try to have your cake and eat it too: let the system build its own symbols, with attached parameters that 'mean' whatever they end up meaning after the symbols have been built, BUT then at the same time insist that some of the parameters really do 'mean' things like probabilities or likelihood or confidence values. If the programmer does anything at all to include mechanisms that rely on these meanings (these interpretations of what the parameters signify) then the programmer has second-guessed what the system itself was going to use those things for, and you have a conflict between the two. My strategy is to keep my hands off, not do anything to strictly interpret those parameters, and experimentally observe the properties of systems that seem loosely consistent with the known architecture of human cognition. I have a parameter, for instance, that seems to be a happiness or consistency parameter attached to a knowledge-atom. But beyond roughly characterising it as such, I do not insert any mechanisms that (implicitly or explicitly) lock the system into such an intepretation. Instead, I have a wide variety of different candidate mechanisms that use that parameter, and I look at the overall properties of systems that use these different candidate mechanisms. I let the system use the parameter according to the dictates of whatever mechanism is in place, but then I just explore the consequences (the high level behavior of the system). In this way I do not get a conflict between what I think the parameter 'ought' to mean and what the system is implicitly taking it to 'mean' by its use of the parameter. I could start talking about all the different candidate mechanisms, but there are thousands of them (at least thousands of candidates that I go so far as to test: they are generated in a semi-automatic way, so there are an unlimited number of potential candidates). 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/?member_id=8660244id_secret=65198894-3ece99
[agi] Polyworld: Using Evolution to Design Artificial Intelligence
This may be of interest to the group. http://video.google.com/videoplay?docid=-112735133685472483 This presentation is about a potential shortcut to artificial intelligence by trading mind-design for world-design using artificial evolution. Evolutionary algorithms are a pump for turning CPU cycles into brain designs. With exponentially increasing CPU cycles while our understanding of intelligence is almost a flat-line, the evolutionary route to AI is a centerpiece of most Kurzweilian singularity scenarios. This talk introduces the Polyworld artificial life simulator as well as results from our ongoing attempt to evolve artificial intelligence and further the Singularity. Polyworld is the brain child of Apple Computer Distinguished Scientist Larry Yaeger, who remains the primary developer of Polyworld: http://www.beanblossom.in.us/larryy/P... Speaker: Virgil Griffith Virgil Griffith is a first year graduate student in Computation and Neural Systems at the California Institute of Technology. On weekdays he studies evolution, computational neuroscience, and artificial life. He did computer security work until his first year of university when his work got him sued for sedition and espionage. He then decided that security was probably not safest field to be in and he turned his life to science. (less) Added: November 13, 2007 - Jef - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244id_secret=65277465-3d25ea
Re: [agi] What best evidence for fast AI?
Bryan Bishop wrote: On Wednesday 14 November 2007 11:55, Richard Loosemore wrote: I was really thinking of the data collection problem: we cannot take one brain and get full information about all those things, down to a sufficient level of detail. I do not see such a technology even over the horizon (short of full-blow nanotechnology) that can deliver that. We can get different information from different individual brains (all of them dead), but combining that would not necessarily be meaningful: all brains are different. Re: all brains are different. What about the possibilities of cloning mice and then proceeding to raise them in Skinner boxes with the exact same environmental conditions, the same stimulation routines, etc. ? Ideally this will give us a baseline mouse that is not only genetically similar, but also behaviorally similar to some degree. This would undoubtedly be helpful in this quest. Well, now you have suggested this I am sure some neuroscientist will do it ;-). But you have to understand that I am a cognitive scientist, with a huge agenda that involves making good use of what I see as the uneplxored fertile ground between cognitive science and AI and I think that I will be able to build an AGI using this approach *long* before the neuroscientists even get one mouse-brain scan at the neuron level (never mind the synaptic bouton level)! So: yeah, but not necessary. 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/?member_id=8660244id_secret=65204588-4868d1
