[agi] Grounding
Okay, I am bored, or maybe just lazy today, so please let me weigh in and ramble a bit: Vectors and scalars are great, and may be the best route to learning for a given system, but it hardly seems obvious that they are a prerequisite to learning for an AI that exceeds general human intellectual capacity. I was a chemical engineer in one of my former lives, and I can say that vectors are definitely more lovable than the criminal defendants I was appointed to represent in my former life as an attorney. The defendants were mostly interested in the rather binary guilty vs. not guilty. Retinas have pixels don't they? Perhaps our perception of scalars is actually recognition of patterns in discrete points. You could readily make an image people recognize as a circle, using only pawns as discrete points on a chessboard. Wouldn't chess be a domain where an AGI could learn and excel, with no vectors or scalars in sight? Much of what is fundamental is binary: on/off, dead/alive, male/female, married/single, smile/frown, and so on. A miss is a good as a mile. . . . Kevin C. P.S. To me a key fundamental is "Artificial Motivation." Give an entity the desire to accomplish goals, plus tools to use, then the ability to learn. Example: I was hungry, but now am full. I wanted to reproduce, and satisfied that urge. Now I am tired of thinking, and want to consume more of that wet fermented grain to stop the process for a while. Ahh, cultivating barely to make beer is good. Oops, inadvertently founded civilization. --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] AI on TV
On 12/9/02 7:13 PM, "Pei Wang" <[EMAIL PROTECTED]> wrote: > On this issue, we can distinguish 4 approaches: > > (1) let symbols get their meaning through "interpretation" (provided in > another language) --- this is the approach used in traditional symbolic AI. > > (2) let symbols get their meaning by grounding on textual experience --- > this is what I and Kevin suggested. > > (3) let symbols get their meaning by grounding on simplified perceptual > experience --- this is what Ben and Shane suggested. > > (4) let symbols get their meaning by grounding on human-level perceptual > experience --- this is what Brooks (the robotics researcher at MIT) and > Harnad (who raised the "symbol grounding" issue in the first place) > proposed. I can be put pretty much in the (2) camp. This is adequate for proving the basic capability of the system and you can incrementally add (3+) later. I mostly view this as a pragmatic engineering issue though; no need to unnecessarily complicate the test environment until you can prove the system is capable of handling the simplest environment. It is a much easier development trajectory unless you believe that (3) or (4) are an absolute minimum for the system to work at all (obviously I don't). Cheers, -James Rogers [EMAIL PROTECTED] --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] general patterns induction
On 12/9/02 7:33 PM, "Pablo" <[EMAIL PROTECTED]> wrote: > > If Solomonoff is powerful enough, I hope it "realizes" by itself that > data is grouped in "words" when it happens so - haha. I'm not working > with human language, but who knows, maybe I'll get in a similar way - > I'll tell you if that happens =) Words will in fact become emergent entities using adaptations of this construct with some time/training, and with even more time/training, sentence structure will also start to emerge. It doesn't even matter what language you are using. Interactive training isn't even necessary unless you want to reduce the number of iterations. I will warn you ahead of time though, that the real trick is the design of the actual software. All data structures/algorithms described in literature thus far are naive and don't scale at all to even vaguely interesting levels of complexity. It can be done, but you'll have to do some original work to make it viable. :-) BTW, although the Li/Vitanyi book is expensive (I get all my Springer-Verlag stuff for free), it is heavily referenced in other papers on the subject and is designed as a graduate-level textbook, which makes it both accessible and a good reference. If you are serious about pursuing this line of thought, it is about as good a book as you'll find on that area of mathematics. It isn't deeply speculative, but it gives you the foundations for everything you'll need to know so that YOU can be deeply speculative. ;-) I don't use my copy much any more, but it is a good bible to have around. Cheers, -James Rogers [EMAIL PROTECTED] --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
RE: [agi] Grounding
True. The more fundamental point is that symbols representing entities and concepts need to be grounded with (scalar) attributes of some sort. How this is *implemented* is a practical matter. One important consideration for AGI is that data is easily retrievable by vector distance (similarity) and that new patterns can be leaned (unlearned) incrementally. Peter http://adaptiveai.com/ -Original Message- Behalf Of Ben Goertzel Well, the fact that clustering requires vectors for A2I2, is a property of your particular AI algorithms... Our Novamente clustering MindAgent is based on the Bioclust clustering algorithm, which does not act on vectors: ... Translating textual experience directly into weighted graphs is often more natural than translating it into vectors. A lot of NLP frameworks use graph representations --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
RE: [agi] Grounding
Well, the fact that clustering requires vectors for A2I2, is a property of your particular AI algorithms... Our Novamente clustering MindAgent is based on the Bioclust clustering algorithm, which does not act on vectors: http://www.math.tau.ac.il/~rshamir/algmb/00/scribe00/html/lec12/node1.html Rather, it acts on (undirected) weighted graphs [which exist as subsets of Novamente's directed weighted hypergraph knowledge representation]. You can always turn a set of vectors into a weighted graph, or vice versa, but the transformation can be very impractical sometimes... Translating textual experience directly into weighted graphs is often more natural than translating it into vectors. A lot of NLP frameworks use graph representations -- Ben > -Original Message- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On > Behalf Of Peter Voss > Sent: Monday, December 09, 2002 10:04 PM > To: [EMAIL PROTECTED] > Subject: [agi] Grounding > > > I think it's more than a matter of 'pragmatics': In order to do > unsupervised > learning (clustering) of grounded entities and concepts, they *must* be > derived from vector-encodable input data. Obviously, not all > inputs need to > represent continuous attributes/ features, but foundational ones do. > > Peter > > http://adaptiveai.com/ > > > > > -Original Message- > Behalf Of Ben Goertzel > > Kevin, > > I'm sure you're right in a theoretical sense, but in practice, I have a > strong feeling it will be a lot easier to teach an AGI stuff if one has a > nonlinguistic world to communicate to it about. > > Rather than just communicating in math and English, I think > teaching will be > much easier if the system can at least perceive 2D pixel > patterns. It'll be > a lot nicer to be able to tell it "There's a circle" when there's a circle > on the screen [that you and it both see] -- to tell it "the > circle is moving > fast", "You stopped the circle", etc. etc. Then to have it see a > whole lot > of circles so that, in an unsupervised way, it gets used to perceiving > them > > This is not a matter of principle, it's a matter of > pragmatics I think > that a perceptual-motor domain in which a variety of cognitively simple > patterns are simply expressed, will make world-grounded early language > learning much easier... > > -- Ben > > --- > To unsubscribe, change your address, or temporarily deactivate > your subscription, > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] > --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
RE: [agi] Tony's 2d World
Tony's 2D training world is a lot simpler than A2I2's, for now. [He is quite free to share details with you or this list, though.] For one thing, his initial shape-world is perception only, involving no action! The simple stuff that we're going to test with it right now, does not involve action, and no advanced perception either, mostly just some aspects of cognition. We're going to meet in January to discuss the details of incorporating action into the shape-world (among other things)... I think that we'll spend some months playing around with very simple prototype worlds. Then we will want to have a really high-quality flexible shape-world framework, and at that point, it would be really grand to coordinate efforts with A2I2 and possibly other AGI projects as well. If Tony proceeds rapidly, that may be around mid-2003 I think that it would be great to have a common testing framework for AGI system, involving 2D shape perception & manipulation, and the option simultaneous NL communication This would save AGI developers work, and would also provide a nice way to compare AGI systems with each other, and to allow various baby AGI systems to interact. -- Ben G > -Original Message- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On > Behalf Of Peter Voss > Sent: Monday, December 09, 2002 10:15 PM > To: [EMAIL PROTECTED] > Subject: [agi] Tony's 2d World > > > Hey Tony - are you on this list? How are you doing? Can we have a look at > your world (or spec)? Perhaps we can co-ordinate our efforts somehow. > > > Peter > > http://adaptiveai.com/ > > > > -Original Message- > Behalf Of Ben Goertzel > > > ... [Although, in fact, Tony Lofthouse is coding up a simple 2D > training-world right now, just to test > some of the current Novamente cognitive functions in isolation, > even though > the system is not yet ready for real experiential learning] > > --- > To unsubscribe, change your address, or temporarily deactivate > your subscription, > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] > --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
RE: [agi] AI on TV
> On this issue, we can distinguish 4 approaches: > > (1) let symbols get their meaning through "interpretation" (provided in > another language) --- this is the approach used in traditional > symbolic AI. > > (2) let symbols get their meaning by grounding on textual experience --- > this is what I and Kevin suggested. > > (3) let symbols get their meaning by grounding on simplified perceptual > experience --- this is what Ben and Shane suggested. > > (4) let symbols get their meaning by grounding on human-level perceptual > experience --- this is what Brooks (the robotics researcher at MIT) and > Harnad (who raised the "symbol grounding" issue in the first place) > proposed. In Novamente, we plan to start with 3 but to fairly quickly move to a combination of 3 and 2 I think that Peter Voss plans to stay with 3 alone for a longer period... -- Ben --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
RE: [agi] general patterns induction
Alan, thanks for the support!!! Your kind words really encourage me on my work =) From now I'll say "pattern discovery" James: I've just placed an order for that book at amazon =) I'll se what I can get from it (I hope a lot, because the book is quite expensive!!) Gary: your book looks interesting too, but I feel the "Solomonoff" thing seems more generic. I'll try your book after that one. > Your project probably involves doing the prediction on a character rather > than a word basis but if you happen to be thinking along the line of words > instead of characters, I would be interested in hearing more about your > work. If Solomonoff is powerful enough, I hope it "realizes" by itself that data is grouped in "words" when it happens so - haha. I'm not working with human language, but who knows, maybe I'll get in a similar way - I'll tell you if that happens =) Cliff: what you told us about the languages was quite interesting. Unfortunately for me, Trigrams is a given pattern; and what I want is the machine to discover patterns by itself. Anyway it's a good approach. It's always good to have handy simple tools when everything gets fuzzy... Thank you all again Kind Regards Pablo Carbonell -Mensaje original- De: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] En nombre de Alan Grimes Enviado el: Lunes, 09 de Diciembre de 2002 02:56 a.m. Para: [EMAIL PROTECTED] Asunto: Re: [agi] general patterns induction Congrats! You have just earned a spot in my list of the world's top-ten AI researchers. Compared to the NLPers, you are the one who is _REALLY_ working on AI. I don't know what the NLPer's are smoking but you are on the right track. I generally frown on the term "pattern recognition" but, perhaps, pattern discovery would work... I wrote an article on my thinking on this reciently that I would have to dig out of the archives of [EMAIL PROTECTED] to send you. Please consider me at your disposal for any additional help on the subject. There are no books on the subject, you are writing your own... Welcome to the cutting edge! =) Pablo wrote: > I'm looking for information about "pattern induction" or "general > patterns" or anything that sounds like that... > > What I want to do is, having a stream of data, predict what may come. > (yes, and then take over the world... sorry if it sounds like Pinky and > The Brain!!) > > I guess general patterns induction is related to data compression, > because if we find a pattern in a string, then we don't have to write > all the characters every time the pattern appears. Surely someone has > already been working on that (who?) > > Anyone would please give me a clue? Is there any book I should read?? Is > there any book like "AI basics", "introduction to AI", or "AI for > dummies" that may help before? > > Thanks a lot! > > Pablo Carbonell > > PS: thanks Ben, Kevin and Eliezer for the previous help > > --- > To unsubscribe, change your address, or temporarily deactivate your subscription, > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] -- pain (n): see Linux. http://users.rcn.com/alangrimes/ --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] --- Incoming mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.417 / Virus Database: 233 - Release Date: 08/11/2002 --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
[agi] Grounding
I think it's more than a matter of 'pragmatics': In order to do unsupervised learning (clustering) of grounded entities and concepts, they *must* be derived from vector-encodable input data. Obviously, not all inputs need to represent continuous attributes/ features, but foundational ones do. Peter http://adaptiveai.com/ -Original Message- Behalf Of Ben Goertzel Kevin, I'm sure you're right in a theoretical sense, but in practice, I have a strong feeling it will be a lot easier to teach an AGI stuff if one has a nonlinguistic world to communicate to it about. Rather than just communicating in math and English, I think teaching will be much easier if the system can at least perceive 2D pixel patterns. It'll be a lot nicer to be able to tell it "There's a circle" when there's a circle on the screen [that you and it both see] -- to tell it "the circle is moving fast", "You stopped the circle", etc. etc. Then to have it see a whole lot of circles so that, in an unsupervised way, it gets used to perceiving them This is not a matter of principle, it's a matter of pragmatics I think that a perceptual-motor domain in which a variety of cognitively simple patterns are simply expressed, will make world-grounded early language learning much easier... -- Ben --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
[agi] Tony's 2d World
Hey Tony - are you on this list? How are you doing? Can we have a look at your world (or spec)? Perhaps we can co-ordinate our efforts somehow. Peter http://adaptiveai.com/ -Original Message- Behalf Of Ben Goertzel ... [Although, in fact, Tony Lofthouse is coding up a simple 2D training-world right now, just to test some of the current Novamente cognitive functions in isolation, even though the system is not yet ready for real experiential learning] --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] AI on TV
On this issue, we can distinguish 4 approaches: (1) let symbols get their meaning through "interpretation" (provided in another language) --- this is the approach used in traditional symbolic AI. (2) let symbols get their meaning by grounding on textual experience --- this is what I and Kevin suggested. (3) let symbols get their meaning by grounding on simplified perceptual experience --- this is what Ben and Shane suggested. (4) let symbols get their meaning by grounding on human-level perceptual experience --- this is what Brooks (the robotics researcher at MIT) and Harnad (who raised the "symbol grounding" issue in the first place) proposed. My opinion is: in principle, the approach (1) doesn't work well for AI, while the last 3 approaches are in the same category. Of course, the richer the experience is, the more capable the system will be. However, to actually develop an AGI theory/system, I'd rather start with (2), and leave (3) for the next step, and (4) for the future. Therefore, though I basically agree with what Ben and Shane said, I won't do that in NARS very soon. Pei - Original Message - From: "Shane Legg" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Monday, December 09, 2002 9:44 PM Subject: Re: [agi] AI on TV > > I think my position is similar to Ben's; it's not really what you > ground things in, but rather that you don't expose your limited > little computer brain to an environment that is too complex -- > at least not to start with. Language, even reasonably simple > context free languages, could well be too rich for a baby AI. > Trying to process 3D input is far too complex. Better then to > start with something simple like 2D pixel patterns as Ben suggests. > The A2I2 project by Peter Voss is taking a similar approach. > > Once very simple concepts and relations have been formed at this > level then I would expect an AI to be better able to start dealing > with richer things like basic language using what it learned > previously as a starting point. For example, relating simple > patterns of language that have an immediate and direct relation > to the visual environment to start with and slowly building up > from there. > > Shane > > --- > To unsubscribe, change your address, or temporarily deactivate your subscription, > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] > --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
RE: [agi] AI on TV
> I think my position is similar to Ben's; it's not really what you > ground things in, but rather that you don't expose your limited > little computer brain to an environment that is too complex -- > at least not to start with. Language, even reasonably simple > context free languages, could well be too rich for a baby AI. > Trying to process 3D input is far too complex. Better then to > start with something simple like 2D pixel patterns as Ben suggests. > The A2I2 project by Peter Voss is taking a similar approach. > > Once very simple concepts and relations have been formed at this > level then I would expect an AI to be better able to start dealing > with richer things like basic language using what it learned > previously as a starting point. For example, relating simple > patterns of language that have an immediate and direct relation > to the visual environment to start with and slowly building up > from there. > > Shane As Shane and I know, but everyone on this list may not to I'll say it anyway, Peter Voss and I discussed this a fair bit before he started the A2I2 project I think we each influenced each others' ideas about AI teaching/training a bit, although we came into the dialogue with some fairly similar ideas on the topic in the first place. The big differences between the A2I2 approach and the Novamente approach are: 1) A2I2 is much closer to a neural net approach [involving neural-gas like stuff, and other NN methods as well, some of them innovative], whereas Novamente occupies a middle ground between subsymbolic & symbolic approaches 2) In the A2I2 project, they're starting off right away with trying to teach the system based on perceptual-motor experience in a simple 2D domain. In Novamente, we are deferring this until we have our (more complex) cognitive infrastructure more fully implemented and tested. [Although, in fact, Tony Lofthouse is coding up a simple 2D training-world right now, just to test some of the current Novamente cognitive functions in isolation, even though the system is not yet ready for real experiential learning] -- Ben G --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] AI on TV
I think my position is similar to Ben's; it's not really what you ground things in, but rather that you don't expose your limited little computer brain to an environment that is too complex -- at least not to start with. Language, even reasonably simple context free languages, could well be too rich for a baby AI. Trying to process 3D input is far too complex. Better then to start with something simple like 2D pixel patterns as Ben suggests. The A2I2 project by Peter Voss is taking a similar approach. Once very simple concepts and relations have been formed at this level then I would expect an AI to be better able to start dealing with richer things like basic language using what it learned previously as a starting point. For example, relating simple patterns of language that have an immediate and direct relation to the visual environment to start with and slowly building up from there. Shane --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] AI on TV
Ben Goertzel wrote: > This is not a matter of principle, it's a matter of pragmatics I > think that a perceptual-motor domain in which a variety of cognitively > simple patterns are simply expressed, will make world-grounded early > language learning much easier... If anyone has the software for this, please tell me! =) -- pain (n): see Linux. http://users.rcn.com/alangrimes/ --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
RE: [agi] AI on TV
Kevin, I'm sure you're right in a theoretical sense, but in practice, I have a strong feeling it will be a lot easier to teach an AGI stuff if one has a nonlinguistic world to communicate to it about. Rather than just communicating in math and English, I think teaching will be much easier if the system can at least perceive 2D pixel patterns. It'll be a lot nicer to be able to tell it "There's a circle" when there's a circle on the screen [that you and it both see] -- to tell it "the circle is moving fast", "You stopped the circle", etc. etc. Then to have it see a whole lot of circles so that, in an unsupervised way, it gets used to perceiving them This is not a matter of principle, it's a matter of pragmatics I think that a perceptual-motor domain in which a variety of cognitively simple patterns are simply expressed, will make world-grounded early language learning much easier... -- Ben > -Original Message- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On > Behalf Of maitri > Sent: Monday, December 09, 2002 5:52 PM > To: [EMAIL PROTECTED] > Subject: Re: [agi] AI on TV > > > I don't want to underestimate the value of embodiment for an AI system, > especially for the development of consciousness. But this is just my > opinion... > > As far as a very useful AGI, I don't see the necessity of a body > or sensory > inputs beyond textual input. Almost any form can be represented as > mathematical models that can easily be input to the system in that manner. > I'm sure there are others on this list that have thought a lot more about > this than I have.. > > Kevin > > - Original Message - > From: "Shane Legg" <[EMAIL PROTECTED]> > To: <[EMAIL PROTECTED]> > Sent: Monday, December 09, 2002 4:18 PM > Subject: Re: [agi] AI on TV > > > > Gary Miller wrote: > > > On Dec. 9 Kevin said: > > > > > > "It seems to me that building a strictly "black box" AGI that > only uses > > > text or graphical input\output can have tremendous > implications for our > > > society, even without arms and eyes and ears, etc. Almost > anything can > > > be designed or contemplated within a computer, so the need for dealing > > > with analog input seems unnecessary to me. Eventually, these will be > > > needed to have a complete, human like AI. It may even be better that > > > these first AGI systems will not have vision and hearing > because it will > > > make it more palatable and less threatening to the masses" > > > > My understanding is that this current trend came about as follows: > > > > Classical AI system where either largely disconnected from the physical > > world or lived strictly in artificial mirco worlds. This lead to a > > number of problems including the famous "symbol grounding problem" where > > the agent's symbols lacked any grounding in an external reality. As a > > reaction to these problems many decided that AI agents needed to be > > more grounded in the physical world, "embodiment" as they call it. > > > > Some now take this to an extreme and think that you should start with > > robotic and sensory and control stuff and forget about logic and what > > thinking is and all that sort of thing. This is what you see now in > > many areas of AI research, Brooks and the Cog project at MIT being > > one such example. > > > > Shane > > > > > > --- > > To unsubscribe, change your address, or temporarily deactivate your > subscription, > > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] > > > --- > To unsubscribe, change your address, or temporarily deactivate > your subscription, > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] > --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] AI on TV
I have a paper (http://www.cogsci.indiana.edu/farg/peiwang/PUBLICATION/#semantics) on this topic, which is mostly in agreement with what Kevin said. For an intelligent system, it is important for its concepts and beliefs to be grounded on the system's experience, but such experience can be textual. Of course, sensorimotor experience is richer, but it is not fundamentally different from textual experience. Pei - Original Message - From: "maitri" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Monday, December 09, 2002 5:52 PM Subject: Re: [agi] AI on TV > I don't want to underestimate the value of embodiment for an AI system, > especially for the development of consciousness. But this is just my > opinion... > > As far as a very useful AGI, I don't see the necessity of a body or sensory > inputs beyond textual input. Almost any form can be represented as > mathematical models that can easily be input to the system in that manner. > I'm sure there are others on this list that have thought a lot more about > this than I have.. > > Kevin > > - Original Message - > From: "Shane Legg" <[EMAIL PROTECTED]> > To: <[EMAIL PROTECTED]> > Sent: Monday, December 09, 2002 4:18 PM > Subject: Re: [agi] AI on TV > > > > Gary Miller wrote: > > > On Dec. 9 Kevin said: > > > > > > "It seems to me that building a strictly "black box" AGI that only uses > > > text or graphical input\output can have tremendous implications for our > > > society, even without arms and eyes and ears, etc. Almost anything can > > > be designed or contemplated within a computer, so the need for dealing > > > with analog input seems unnecessary to me. Eventually, these will be > > > needed to have a complete, human like AI. It may even be better that > > > these first AGI systems will not have vision and hearing because it will > > > make it more palatable and less threatening to the masses" > > > > My understanding is that this current trend came about as follows: > > > > Classical AI system where either largely disconnected from the physical > > world or lived strictly in artificial mirco worlds. This lead to a > > number of problems including the famous "symbol grounding problem" where > > the agent's symbols lacked any grounding in an external reality. As a > > reaction to these problems many decided that AI agents needed to be > > more grounded in the physical world, "embodiment" as they call it. > > > > Some now take this to an extreme and think that you should start with > > robotic and sensory and control stuff and forget about logic and what > > thinking is and all that sort of thing. This is what you see now in > > many areas of AI research, Brooks and the Cog project at MIT being > > one such example. > > > > Shane > > > > > > --- > > To unsubscribe, change your address, or temporarily deactivate your > subscription, > > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] > > > --- > To unsubscribe, change your address, or temporarily deactivate your subscription, > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] > --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] AI on TV
I don't want to underestimate the value of embodiment for an AI system, especially for the development of consciousness. But this is just my opinion... As far as a very useful AGI, I don't see the necessity of a body or sensory inputs beyond textual input. Almost any form can be represented as mathematical models that can easily be input to the system in that manner. I'm sure there are others on this list that have thought a lot more about this than I have.. Kevin - Original Message - From: "Shane Legg" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Monday, December 09, 2002 4:18 PM Subject: Re: [agi] AI on TV > Gary Miller wrote: > > On Dec. 9 Kevin said: > > > > "It seems to me that building a strictly "black box" AGI that only uses > > text or graphical input\output can have tremendous implications for our > > society, even without arms and eyes and ears, etc. Almost anything can > > be designed or contemplated within a computer, so the need for dealing > > with analog input seems unnecessary to me. Eventually, these will be > > needed to have a complete, human like AI. It may even be better that > > these first AGI systems will not have vision and hearing because it will > > make it more palatable and less threatening to the masses" > > My understanding is that this current trend came about as follows: > > Classical AI system where either largely disconnected from the physical > world or lived strictly in artificial mirco worlds. This lead to a > number of problems including the famous "symbol grounding problem" where > the agent's symbols lacked any grounding in an external reality. As a > reaction to these problems many decided that AI agents needed to be > more grounded in the physical world, "embodiment" as they call it. > > Some now take this to an extreme and think that you should start with > robotic and sensory and control stuff and forget about logic and what > thinking is and all that sort of thing. This is what you see now in > many areas of AI research, Brooks and the Cog project at MIT being > one such example. > > Shane > > > --- > To unsubscribe, change your address, or temporarily deactivate your subscription, > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] AI on TV
that's him... - Original Message - From: "Shane Legg" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Monday, December 09, 2002 3:43 PM Subject: Re: [agi] AI on TV > maitri wrote: > > > > The second guy was from either England or the states, not sure. He was > > working out of his garage with his wife. He was trying to develop robot > > AI including vision, speech, hearing and movement. > > This one's a bit more difficult, Steve Grand perhaps? > > http://www.cyberlife-research.com/people/steve/ > > Shane > > --- > To unsubscribe, change your address, or temporarily deactivate your subscription, > please go to http://v2.listbox.com/member/?[EMAIL PROTECTED] > --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] AI on TV
Gary Miller wrote: On Dec. 9 Kevin said: "It seems to me that building a strictly "black box" AGI that only uses text or graphical input\output can have tremendous implications for our society, even without arms and eyes and ears, etc. Almost anything can be designed or contemplated within a computer, so the need for dealing with analog input seems unnecessary to me. Eventually, these will be needed to have a complete, human like AI. It may even be better that these first AGI systems will not have vision and hearing because it will make it more palatable and less threatening to the masses" My understanding is that this current trend came about as follows: Classical AI system where either largely disconnected from the physical world or lived strictly in artificial mirco worlds. This lead to a number of problems including the famous "symbol grounding problem" where the agent's symbols lacked any grounding in an external reality. As a reaction to these problems many decided that AI agents needed to be more grounded in the physical world, "embodiment" as they call it. Some now take this to an extreme and think that you should start with robotic and sensory and control stuff and forget about logic and what thinking is and all that sort of thing. This is what you see now in many areas of AI research, Brooks and the Cog project at MIT being one such example. Shane --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] AI on TV
maitri wrote: The second guy was from either England or the states, not sure. He was working out of his garage with his wife. He was trying to develop robot AI including vision, speech, hearing and movement. This one's a bit more difficult, Steve Grand perhaps? http://www.cyberlife-research.com/people/steve/ Shane --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
RE: [agi] AI on TV
I was at Starlab one week after it folded. Hugo was the only one left there -- he was living in an apartment in the building. It was a huge, beautiful, ancient, building, formerly the Czech Embassy to Brussels I saw the CAM-Brain machine (CBM) there, disabled by Korkin (the maker) due to non-payment... There is a CBM in use at ATR in Japan [where Hugo used to work], but it's mostly being used for simple hardware-type experiments, not advanced learning... ; there was one at Lernout-Hauspie, but I don't know what became of it when that firm went under... Hugo is currently designing the CBM-2, and I've given him some possibly useful ideas in that regard... I can sympathize somewhat with Korkin: he spent his own $$ on the hardware, and then starlab did not pay him, breaking its contractual obligations. He is struggling financially. And Hugo was not at all politic or sympathetic in dealing with him, because Hugo is always so wrapped up in his own problems. Well, such is human life I tried briefly to help smooth things over w/ Korkin, but Hugo's attitude was sufficiently out-there that it was not possible... -- Ben -Original Message-From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of maitriSent: Monday, December 09, 2002 11:44 AMTo: [EMAIL PROTECTED]Subject: Re: [agi] AI on TV Ben, I just read the Bio. You gave alot more play to his ideas than the show did. You probably know this, but Starlab has folded and I think he was off to the states... The show seemed to indicate that nothing of note ever came out of the project. In fact, it appeared to not generate one new network . What they didn't detail was the cause of this. It could have ben hardware related, I don't know. They were also having serious contract problems with the Russian fellow who built it. He had effectively disabled the machine from the US until he got some more money, which eventually killed the whole thing. What a waste. Maybe you can buy the machine off Ebay now. They said it would be auctioned... They did give alot of play to his seemingly contrarion ideas about the implications of his work. It was a rather dismal outlook on societies lack of general acceptance of AI and\or enhancement. I hope he was off base in this area, but I wouldn't be surprised if a small group of radical anti-AI people emerge with hostile intent. Another good reason to not be so visible!! Kevin - Original Message - From: Ben Goertzel To: [EMAIL PROTECTED] Sent: Monday, December 09, 2002 11:26 AM Subject: RE: [agi] AI on TV There was a show on the tube last night on TechTV. It was part of their weekly Secret, Strange and True series. They chronicled three guys who are working on creating advanced AI. One guy was from Belgium. My apologies to him if he reads this list, but he was a rather quirky and stressed character. He had designed a computer that was basically a collection of chips. He raised a million and had it built on spec. I gather he was expecting something to miraculously emerge from this collection, but alas, nothing did. It was really stressful watching his stress. He had very high visibility in the country and the pressure was immense as he promised a lot. I have real doubts about his approach, even though I am a lay-AI person. Also, its clear from watching him that its sometimes good to have shoestring budgets and low visibility. Less stress and more forced creativity in your approach... Kevin: Was the guy from Belgium perhaps Hugo de Garis?? [Who is not in Belgium anymore, but who designed a radical hardware based approach to AGI, and who is a bit of a quirky guy?? ...] I visited Hugo at Starlab [when it existed] in Brussels in mid-2001 See my brief bio of Hugo at http://www.goertzel.org/benzine/deGaris.htm -- Ben G
RE: [agi] AI on TV
Title: Message On Dec. 9 Kevin said: "It seems to me that building a strictly "black box" AGI that only uses text or graphical input\output can have tremendous implications for our society, even without arms and eyes and ears, etc. Almost anything can be designed or contemplated within a computer, so the need for dealing with analog input seems unnecessary to me. Eventually, these will be needed to have a complete, human like AI. It may even be better that these first AGI systems will not have vision and hearing because it will make it more palatable and less threatening to the masses" I agree wholeheartedly. Sony and Honda as well as several military contractors are spending 10s perhaps hundreds of million dollars on R&D robotics programs which incorporate the vision, and analog control, and data acquisition for industry, the military, and yes even the toy companies. Once AGIs are ready to fly it will be able to interface with these systems through software APIs (Application Programming Interfaces) and will not even care about the low-level programs that enable them move about and visually survey their environments. Too often those who seek the spotlight are really sincere, but either need recognition for their own self reassurance or as a method of attracting potential funding. There seems to be an unwritten law in the universe which that says all major inventions will involve major sacrifice and loss for those who dare to tackle what has been deemed impossible by others. From Galileo to Edison, to Tesla, to maybe one of us. Before we succeed, if we succeed, the universe will exact it's toll. For nature will not give up her secrets willingly and intelligence may be her most closely guarded secret of all! Don't forget that genius and madness sometimes walk arm in arm! And as the man says if you weren't cazy when you got in, you probably will be before you get out!. -Original Message-From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] On Behalf Of maitriSent: Monday, December 09, 2002 11:08 AMTo: [EMAIL PROTECTED]Subject: [agi] AI on TV There was a show on the tube last night on TechTV. It was part of their weekly Secret, Strange and True series. They chronicled three guys who are working on creating advanced AI. One guy was from Belgium. My apologies to him if he reads this list, but he was a rather quirky and stressed character. He had designed a computer that was basically a collection of chips. He raised a million and had it built on spec. I gather he was expecting something to miraculously emerge from this collection, but alas, nothing did. It was really stressful watching his stress. He had very high visibility in the country and the pressure was immense as he promised a lot. I have real doubts about his approach, even though I am a lay-AI person. Also, its clear from watching him that its sometimes good to have shoestring budgets and low visibility. Less stress and more forced creativity in your approach... The second guy was from either England or the states, not sure. He was working out of his garage with his wife. He was trying to develop robot AI including vision, speech, hearing and movement. He was clearly floundering as he radically redesigned what he was doing probably a dozen times during the 1 hour show. I think this experimentation has value. But I really wonder if large scale trial and error will result in AGI. I don't think so. I think trial and error will, of course, be essential during development, but T and E of the entire underlying architecture seems a folly to me. Since the problem is SO immense, I believe one must start with a very sound and detailed game plan that can be tweaked as things move along. The last guy was brooks at MIT. They were developing a robot with enhanced vision capabilities. They also failed miserably. I am rather glad that they did. They re funded by DOD, and are basically trying to build a robotic killing machine. Just what we need. It seems to me that trying to tackle the vision problem is too big of a place to start. While all this work will have value down the line, is it essential to AGI? It seems to me that building a strictly "black box" AGI that only uses text or graphical input\output can have tremendous implications for our society, even without arms and eyes and ears, etc. Almost anything can be designed or contemplated within a computer, so the need for dealing with analog input seems unnecessary to me. Eventually, these will be needed to have a complete, human like AI. It may even be better that these first AGI systems will not have vision and hearing because it will make it more palatable and less threatening to the masses The show was rather discouraging, especially if one consider
Re: [agi] AI on TV
Ben, I just read the Bio. You gave alot more play to his ideas than the show did. You probably know this, but Starlab has folded and I think he was off to the states... The show seemed to indicate that nothing of note ever came out of the project. In fact, it appeared to not generate one new network . What they didn't detail was the cause of this. It could have ben hardware related, I don't know. They were also having serious contract problems with the Russian fellow who built it. He had effectively disabled the machine from the US until he got some more money, which eventually killed the whole thing. What a waste. Maybe you can buy the machine off Ebay now. They said it would be auctioned... They did give alot of play to his seemingly contrarion ideas about the implications of his work. It was a rather dismal outlook on societies lack of general acceptance of AI and\or enhancement. I hope he was off base in this area, but I wouldn't be surprised if a small group of radical anti-AI people emerge with hostile intent. Another good reason to not be so visible!! Kevin - Original Message - From: Ben Goertzel To: [EMAIL PROTECTED] Sent: Monday, December 09, 2002 11:26 AM Subject: RE: [agi] AI on TV There was a show on the tube last night on TechTV. It was part of their weekly Secret, Strange and True series. They chronicled three guys who are working on creating advanced AI. One guy was from Belgium. My apologies to him if he reads this list, but he was a rather quirky and stressed character. He had designed a computer that was basically a collection of chips. He raised a million and had it built on spec. I gather he was expecting something to miraculously emerge from this collection, but alas, nothing did. It was really stressful watching his stress. He had very high visibility in the country and the pressure was immense as he promised a lot. I have real doubts about his approach, even though I am a lay-AI person. Also, its clear from watching him that its sometimes good to have shoestring budgets and low visibility. Less stress and more forced creativity in your approach... Kevin: Was the guy from Belgium perhaps Hugo de Garis?? [Who is not in Belgium anymore, but who designed a radical hardware based approach to AGI, and who is a bit of a quirky guy?? ...] I visited Hugo at Starlab [when it existed] in Brussels in mid-2001 See my brief bio of Hugo at http://www.goertzel.org/benzine/deGaris.htm -- Ben G
Re: [agi] AI on TV
Indeed it was... I'll read the bio with interest... - Original Message - From: Ben Goertzel To: [EMAIL PROTECTED] Sent: Monday, December 09, 2002 11:26 AM Subject: RE: [agi] AI on TV There was a show on the tube last night on TechTV. It was part of their weekly Secret, Strange and True series. They chronicled three guys who are working on creating advanced AI. One guy was from Belgium. My apologies to him if he reads this list, but he was a rather quirky and stressed character. He had designed a computer that was basically a collection of chips. He raised a million and had it built on spec. I gather he was expecting something to miraculously emerge from this collection, but alas, nothing did. It was really stressful watching his stress. He had very high visibility in the country and the pressure was immense as he promised a lot. I have real doubts about his approach, even though I am a lay-AI person. Also, its clear from watching him that its sometimes good to have shoestring budgets and low visibility. Less stress and more forced creativity in your approach... Kevin: Was the guy from Belgium perhaps Hugo de Garis?? [Who is not in Belgium anymore, but who designed a radical hardware based approach to AGI, and who is a bit of a quirky guy?? ...] I visited Hugo at Starlab [when it existed] in Brussels in mid-2001 See my brief bio of Hugo at http://www.goertzel.org/benzine/deGaris.htm -- Ben G
RE: [agi] Hello from Kevin Copple
Gary Miller wrote: > I also agree that the AGI approach of modeling and creating a self > learning system is a valid bottom up approach to AGI. But it is much > harder for me with my limited mathematical and conceptual knowledge of > the research to grasp how and when these systems will be able jumpstart > themselves and evolve to the point of communicating in English. Sure. In my view, the path involves teaching an AGI to carry out simple tasks in an environment (physical or digital) and then teaching it to communicate about these tasks and related entities in its environment... > While it is true that most bots today generate a reflexive response > based only on the user's input, it is possible to extend bot technology > by generating the response based upon the following additional internal > stimuli not provided in the current input they are responding to. These > stimuli provide at least a portion of the grounding I think you are > referring to. Hm... Actually, I think you're getting at a deep point here. Potentially, *conversational pragmatics* and *inferred psychology* can be used to ground *semantics*, for a chat bot... For example, suppose there's a pattern of word usage, sentence length, etc., which correlates with humans being angry. The bot can learn to correlate this pattern with the word "angry." It is thus grounding the word "angry" with a nonlinguistic pattern... It may then learn different patterns corresponding to "very angry" versus "slightly angry" .. Suppose there's also a pattern of word usage, sentence length, punctuation use, etc., that corresponds to the emotion of "happy" ... and "very happy" vs. "slightly happy" If it also understands "very long sentence" vs. "slightly long sentence" vs. "not long sentence" [via grounding these in sentence lengths], then it may be able to extrapolate from these examples to form an abstract model of "very"-ness in general... Based on this line of thinking, I have to modify and partially retract my previous statement. If a chat bot is given the ability to study patterns in language usage, such as the ones mentioned above, then it may use these patterns as a "nonlinguistic" domain in which to ground its linguistic knowledge... So, I think that truly intelligent language usage COULD potentially be learned by a chat bot I still think this is trickier than learning it via a more physical-world-ish grounding domain, but it's far from impossible Very interesting point, Gary, thanks!! -- Ben --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
RE: [agi] AI on TV
There was a show on the tube last night on TechTV. It was part of their weekly Secret, Strange and True series. They chronicled three guys who are working on creating advanced AI. One guy was from Belgium. My apologies to him if he reads this list, but he was a rather quirky and stressed character. He had designed a computer that was basically a collection of chips. He raised a million and had it built on spec. I gather he was expecting something to miraculously emerge from this collection, but alas, nothing did. It was really stressful watching his stress. He had very high visibility in the country and the pressure was immense as he promised a lot. I have real doubts about his approach, even though I am a lay-AI person. Also, its clear from watching him that its sometimes good to have shoestring budgets and low visibility. Less stress and more forced creativity in your approach... Kevin: Was the guy from Belgium perhaps Hugo de Garis?? [Who is not in Belgium anymore, but who designed a radical hardware based approach to AGI, and who is a bit of a quirky guy?? ...] I visited Hugo at Starlab [when it existed] in Brussels in mid-2001 See my brief bio of Hugo at http://www.goertzel.org/benzine/deGaris.htm -- Ben G
[agi] AI on TV
There was a show on the tube last night on TechTV. It was part of their weekly Secret, Strange and True series. They chronicled three guys who are working on creating advanced AI. One guy was from Belgium. My apologies to him if he reads this list, but he was a rather quirky and stressed character. He had designed a computer that was basically a collection of chips. He raised a million and had it built on spec. I gather he was expecting something to miraculously emerge from this collection, but alas, nothing did. It was really stressful watching his stress. He had very high visibility in the country and the pressure was immense as he promised a lot. I have real doubts about his approach, even though I am a lay-AI person. Also, its clear from watching him that its sometimes good to have shoestring budgets and low visibility. Less stress and more forced creativity in your approach... The second guy was from either England or the states, not sure. He was working out of his garage with his wife. He was trying to develop robot AI including vision, speech, hearing and movement. He was clearly floundering as he radically redesigned what he was doing probably a dozen times during the 1 hour show. I think this experimentation has value. But I really wonder if large scale trial and error will result in AGI. I don't think so. I think trial and error will, of course, be essential during development, but T and E of the entire underlying architecture seems a folly to me. Since the problem is SO immense, I believe one must start with a very sound and detailed game plan that can be tweaked as things move along. The last guy was brooks at MIT. They were developing a robot with enhanced vision capabilities. They also failed miserably. I am rather glad that they did. They re funded by DOD, and are basically trying to build a robotic killing machine. Just what we need. It seems to me that trying to tackle the vision problem is too big of a place to start. While all this work will have value down the line, is it essential to AGI? It seems to me that building a strictly "black box" AGI that only uses text or graphical input\output can have tremendous implications for our society, even without arms and eyes and ears, etc. Almost anything can be designed or contemplated within a computer, so the need for dealing with analog input seems unnecessary to me. Eventually, these will be needed to have a complete, human like AI. It may even be better that these first AGI systems will not have vision and hearing because it will make it more palatable and less threatening to the masses The show was rather discouraging, especially if one considers that these three folks are leading the way towards AGI. As for me, I think others in the field are alot further along...Nonetheless, I'm sure the show will be rerun and may be a worthwhile watch for those here... Kevin
RE: [agi] Hello from Kevin Copple
Ben you said:
RE: [agi] EllaZ systems
Hey Ben, It seems that recent college IT grads here hope to earn about 3000rmb (375usd) a month, but often must settle for less. This is based on my rather limited knowledge. Hopefully I will know more in the near future, since I have been getting the word out and have a local headhunter looking for some candidates. One prospect who is not willing to leave his job for short term work responded, "you are offering too much." >I guess the important thing is to store as much data as possible, in a >clearly structured way. >People can always postprocess the data using their own scripts, so long as >the information is there are and is clearly structured... Yes, I agree with this sentiment. I am thinking along the lines full conventional citation plus other data such as location and original date of creation. We may indulge in a little overkill, since I have already experienced remorse at not recording more detail in some of the early stages. Trial and error remains a great teacher. >XML or RDF type syntax is generally easy for people to work with... XML may be the way to go. Perhaps XML files can largely replace DB's, and a translation from XML to a DB should be straightforward. A relational DB could allow associating one convun to another, thus illustrating a joke or poem, for example. Those types of relationships may be difficult with XML, but could be done programmatically, at least to some extent. This AI business sure could consume a lot of "gurus." >I would definitely want each conversational unit linked to each conversation >it was embodied in -- the full conversational history... so that the context >could be determined One of the interesting things to mine from this >dataset is how people respond to context... I will add "Ben" to my WordNet gloss for "ambitious" :-) . . . good point though. We are now able to conveniently store mind-boggling amounts of text data. Ella will display the entire text of Kant's Critique of Pure Reason in a single window of your browser (its amazing that those scrollbars never wear out). The one-microprocessor bottleneck is the big limitation (for me anway). >On a different topic: If you plan to involve statistical NLP technology in >the next phase of your project, that could be an interesting thing to talk >about ... it's not something I'm working on now, but we played around with >it a lot at Webmind Inc. ... Thanks for the idea. I have been meaning to take a closer look at what has gone on at Webmind Inc.. Later . . . Kevin --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
[agi] Reverse engineering
Here's a good writeup on a team working to reverse engineer the brain... http://www.discover.com/dec_02/feattech.html Kevin
RE: [agi] EllaZ systems
Hi Kevin, > Since wages are so low > here, even for well-educated people, I am in the process of hiring a few > people for a year or so to move our project along faster. Please let me > know if you have any leads or suggestions. I am curious: How much does it cost, roughly, to hire a good programmer there with the ability to understand AI concepts? [Not that I plan to expand Novamente's software operations to China at the moment, I'm just curious ;) ] > Ben, one of the challenges it seems is how best to structure the Convun > database so as to maximize its use for intelligent systems. > There is likely > no clear correct approach, so we will just do our best. I will > try soon to > submit a description of where we are headed to this mailing list > and ask for > comments. I guess the important thing is to store as much data as possible, in a clearly structured way. People can always postprocess the data using their own scripts, so long as the information is there are and is clearly structured... XML or RDF type syntax is generally easy for people to work with... I would definitely want each conversational unit linked to each conversation it was embodied in -- the full conversational history... so that the context could be determined One of the interesting things to mine from this dataset is how people respond to context... On a different topic: If you plan to involve statistical NLP technology in the next phase of your project, that could be an interesting thing to talk about ... it's not something I'm working on now, but we played around with it a lot at Webmind Inc. ... -- Ben --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]