Re: AW: [agi] How general can be and should be AGI?
Mike Tintner wrote: Charles: Flaws in Hamlet: I don't think of this as involving general intelligence. Specialized intelligence, yes, but if you see general intelligence at work there you'll need to be more explicit for me to understand what you mean. Now determining whether a particular deviation from iambic pentameter was a flaw would require a deep human intelligence, but I don't feel that understanding of how human emotions are structured is a part of general intelligence except on a very strongly superhuman level. The level where the AI's theory of your mind was on a par with, or better than, your own. Charles, My flabber is so ghasted, I don't quite know what to say. Sorry, I've never come across any remarks quite so divorced from psychological reality. There are millions of essays out there on Hamlet, each one of them different. Why don't you look at a few?: http://www.123helpme.com/search.asp?text=hamlet I've looked at a few (though not those). In college I formed the definite impression that essays on the meaning of literature were exercises in determining what the instructor wanted. This isn't something that I consider a part of general intelligence (except as mentioned above). ... The reason over 70 per cent of students procrastinate when writing essays like this about Hamlet, (and the other 20 odd per cent also procrastinate but don't tell the surveys), is in part that it is difficult to know which of the many available approaches to take, and which of the odd thousand lines of text to use as support, and which of innumerable critics to read. And people don't have a neat structure for essay-writing to follow. (And people are inevitably and correctly afraid that it will all take if not forever then far, far too long). The problem is that most, or at least many, of the approaches are defensible, but your grade will be determined by the taste of the instructor. This isn't a problem of general intelligence except at a moderately superhuman level. Human tastes aren't reasonable ingredients for an entry level general intelligence. Making it a requirement merely ensures that one will never be developed (whose development attends to your theories of what's required). ... In short, essay writing is an excellent example of an AGI in action - a mind freely crossing different domains to approach a given subject from many fundamentally different angles. (If any subject tends towards narrow AI, it is normal as opposed to creative maths). I can see story construction as a reasonable goal for an AGI, but at the entry level they are going to need to be extremely simple stories. Remember that the goal structures of the AI won't match yours, so only places where the overlap is maximal are reasonable grounds for story construction. Otherwise this is an area for specialized AIs, which isn't what we are after. Essay writing also epitomises the NORMAL operation of the human mind. When was the last time you tried to - or succeeded in concentrating for any length of time? I have frequently written essays and other similar works. My goal structures, however, are not generalized, but rather are human. I have built into me many special purpose functions for dealing with things like plot structure, family relationships, relative stages of growth, etc. As William James wrote of the normal stream of consciousness: Instead of thoughts of concrete things patiently following one another in a beaten track of habitual suggestion, we have the most abrupt cross-cuts and transitions from one idea to another, the most rarefied abstractions and discriminations, the most unheard-of combinations of elements, the subtlest associations of analogy; in a word, we seem suddenly introduced into a seething caldron of ideas, where everything is fizzling and bobbing about in a state of bewildering activity, where partnerships can be joined or loosened in an instant, treadmill routine is unknown, and the unexpected seems the only law. Ditto: The normal condition of the mind is one of informational disorder: random thoughts chase one another instead of lining up in logical causal sequences. Mihaly Csikszentmihalyi Ditto the Dhammapada, Hard to control, unstable is the mind, ever in quest of delight, When you have a mechanical mind that can a) write essays or tell stories or hold conversations [which all present the same basic difficulties] and b) has a fraction of the difficulty concentrating that the brain does and therefore c) a fraction of the flexibility in crossing domains, then you might have something that actually is an AGI. You seem to be placing an extremely high bar in place before you will consider something an AGI. Accepting all that you have said, for an AGI to react as a human would react would require that the AGI be strongly superhuman. More to the point, I wouldn't DARE create an AGI which had motivations similar to
AW: AW: AW: [agi] How general can be and should be AGI?
Matt Mahoney [mailto:[EMAIL PROTECTED] wrote Object oriented programming is good for organizing software but I don't think for organizing human knowledge. It is a very rough approximation. We have used O-O for designing ontologies and expert systems (IS-A links, etc), but this approach does not scale well and does not allow for incremental learning from examples. It totally does not work for language modeling, which is the first problem that AI must solve. I agree that the O-O paradigm is not adequate to model all learning algorithms and models we use. My own example of recognizing voices should show that I have doubts that we use O-O models in our brain for everything of our environment. I think our brain learns a somewhat a hierarchical model of the world. And the algorithm for the low level (e.g. voices, sounds) are probably complete different from the algorithms for higher levels of our models. It is evident that a child has learning capabilities that are far beyond those from an adult. The reason is not only that the child's brain is nearly empty. The physiological architecture is different to some degree. So we can expect that learning the basic low levels of a world model requires algorithms which we only have had as a child. And the result of that learning is to some degree used for bias in later learning algorithm when we are adult. For example we had to learn to extract syllables from the sound wave of spoken language. Learning the grammar rules are in higher levels. Learning semantics is still higher and so on. But it is a matter of fact that we use an O-O like model in the top-levels of our world. You can see this also from language grammar. Subjects objects, predicates, adjectives have their counterparts in the O-O paradigm. A photo of a certain scene is physically an array of colored pixels. But you can ask a human what he sees. And a possible answer could be: Well, there is a house. A man walks to the door. It wears a blue shirt. A woman looks through the window ... Obviously, the answer shows a lot how people model the world in their top-level (= conscious) And obviously the model consists of interacting objects with attributes and behavior. So knowledge representation at higher levels is indeed O-O like. I think your and my answer show that we do not use a single algorithm which is responsible to extract all the regularities from our perceptions. And more important: There is physiological and psychological evidence that the algorithms we use change to some degree during the first decade of our life. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: AW: [agi] How general can be and should be AGI?
Charles, We're still a few million miles apart :). But perhaps we can focus on something constructive here. On the one hand, while, yes, I'm talking about extremely sophisticated behaviour in essaywriting, it has generalizable features that characterise all life. (And I think BTW that a dog is still extremely sophisticated in its motivations and behaviour - your idea there strikes me as evolutionarily naive). Even if a student has an extremely dictatorial instructor, following his instructions slavishly, will be, when you analyse it, a highly problematic, open-ended affair, and no slavish matter - i.e. how he is to apply some general, say, deconstructionist criticism instructions and principles and translate them into a v. complex essay. In fact, it immediately strikes me such essaywriting, and all essaywriting, and most human activities and animal activities will be a matter of hierarchical goals - of, off the cuff, something v. crudely like - write an essay on Hamlet - decide general approach... use deconstructionist approach - find contradictory values in Hamlet to deconstruct...etc. But all life, I guess, must be organized along those lines - the simplest worm must start with something crudely like : find food to eat...decide where food may be located decide approach to food location etc.. (which in turn will almost always be conflicting with opposed emotions/motivations/goals like get some more sleep ..stay cuddled up in burrow.. ) And even, pace Koestler and others, v. simple actions, like reaching out for food in a kitchen, can be a hierarchical affair, with only the general direction and goal decided to begin with, and more specific targeting of arm and shaping of hand, only specified at later stages of the action. Hierarchical goals are surely fundamental to general intelligence. Interestingly, when I Google hierarchical goals and AI, I get v. little - except from our immediate friends, gamers - and this from: Programming Game AI by Example Mat Buckland: Chapter 9: Hierarchical Goal Based Agents This chapter introduces agents that are motivated by hierarchical goals. This type of architecture is far more flexible than the one described in Chapter 2 allowing AI programmers to easily imbue game characters with the brains necessary to do all sorts of funky stuff. Discussion, code and demos of: atomic goals, composite goals, goal arbitration, creating goal evaluation functions, implementation in Raven, using goal evaluations to create personalities, goals and agent memory, automatic resuming of interrupted activities, negotiating special path obstacles such as elevators, doors or moving platforms, command queuing, scripting behavior. Anyone care to comment about using hierarchical goals in AGI or elsewhere? Charles: Flaws in Hamlet: I don't think of this as involving general intelligence. Specialized intelligence, yes, but if you see general intelligence at work there you'll need to be more explicit for me to understand what you mean. Now determining whether a particular deviation from iambic pentameter was a flaw would require a deep human intelligence, but I don't feel that understanding of how human emotions are structured is a part of general intelligence except on a very strongly superhuman level. The level where the AI's theory of your mind was on a par with, or better than, your own. Charles, My flabber is so ghasted, I don't quite know what to say. Sorry, I've never come across any remarks quite so divorced from psychological reality. There are millions of essays out there on Hamlet, each one of them different. Why don't you look at a few?: http://www.123helpme.com/search.asp?text=hamlet I've looked at a few (though not those). In college I formed the definite impression that essays on the meaning of literature were exercises in determining what the instructor wanted. This isn't something that I consider a part of general intelligence (except as mentioned above). ... The reason over 70 per cent of students procrastinate when writing essays like this about Hamlet, (and the other 20 odd per cent also procrastinate but don't tell the surveys), is in part that it is difficult to know which of the many available approaches to take, and which of the odd thousand lines of text to use as support, and which of innumerable critics to read. And people don't have a neat structure for essay-writing to follow. (And people are inevitably and correctly afraid that it will all take if not forever then far, far too long). . This isn't a problem of general intelligence except at a moderately superhuman level. Human tastes aren't reasonable ingredients for an entry level general intelligence. Making it a requirement merely ensures that one will never be developed (whose development attends to your theories of what's required). ... In short, essay writing is an excellent example of an AGI in action - a mind
Language learning (was Re: AW: AW: AW: AW: [agi] How general can be and should be AGI?)
--- Dr. Matthias Heger [EMAIL PROTECTED] wrote: Matt Mahoney [mailto:[EMAIL PROTECTED] wrote Actually that's only true in artificial languages. Children learn words with semantic content like ball and milk before they learn function words like the and of, in spite of their higher frequency. Before they learn the words and their meanings they have to learn to recognize the sounds for the words. And even if they use words like with of and the later they must be able to separate these function-words and relation-words from object-words before they learn any word. But separating words means classifying words and that means knowledge of grammar for a certain degree. Lexical segmentation is learned before semantics, but other grammar is learned afterwards. Babies learn to segment continuous speech into words at 7-10 months [1]. This is before they learn their first word, but is detectable because babies will turn their heads in preference to segmentable speech. It is also possible to guess word divisions in text without spaces given only a statistical knowledge of letter n-grams [2]. Natural language has a structure that makes it easy to learn incrementally from examples with a sufficiently powerful neural network. It must, because any unlearnable features will disappear. Matt Mahoney [mailto:[EMAIL PROTECTED] wrote Techniques for parsing artificial languages fail for natural languages because the parse depends on the meanings of the words, as in the following example: - I ate pizza with pepperoni. - I ate pizza with a fork. - I ate pizza with a friend. In days of early AI the O-O paradigm was not so sophisticated as it is today. The phenomenon of your example is well-known in O-O paradigm and is modeled by overwritten functions which means that Objects may have several functions with the same name but with different signatures. eat(Food f) eat(Food f, ListSideDish l) eat (Food f, ListTool l) eat (Food f, ListPeople l) ... This type of knowledge representation has been tried and it leads to a morass of rules and no intuition on how children learn grammar. We do not know how many grammar rules there are, but it probably exceeds the number of words in our vocabulary, given how long it takes to learn. I think, it is clear that there are representations like classes, objects, relation between objects, attributes of objects. But the crucial questions are: How did we and do we build our O-O models? How created the brain abstract concepts like ball and milk? How do we find classes, objects and relations? We need to understand how children learn grammar without any concept of what a noun or a verb is. Also, how do people learn hierarchical relationships before they learn what a hierarchy is? 1. Jusczyk, Peter W. (1996), Investigations of the word segmentation abilities of infants, 4'th Intl. Conf. on Speech and Language Processing, Vol. 3, 1561-1564. 2. http://cs.fit.edu/~mmahoney/dissertation/lex1.html -- Matt Mahoney, [EMAIL PROTECTED] --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
AW: Language learning (was Re: AW: AW: AW: AW: [agi] How general can be and should be AGI?)
Matt Mahoney [mailto:[EMAIL PROTECTED] wrote eat(Food f) eat(Food f, ListSideDish l) eat (Food f, ListTool l) eat (Food f, ListPeople l) ... This type of knowledge representation has been tried and it leads to a morass of rules and no intuition on how children learn grammar. We do not know how many grammar rules there are, but it probably exceeds the number of words in our vocabulary, given how long it takes to learn. As I said, my intention is not to find a set of O-O like rules to create AGI. The fact that early approaches failed to build AGI by a set of similar rules does not prove, that AGI cannot consist of such rules. For example, there were also approaches to create AI by biological inspired neural networks with some minor success but there was not the real breakthrough too. So this does not prove anything but that the problem of AGI is not so easy to solve. The brain is still a black box regarding many phenomenon. We can analyze our own conscious thoughts and our communication which is nothing else than sending ideas and thoughts from one brain to the other brain via natural language. I am convinced, that the structure and contents of our language is not independent of the internal representation of knowledge. And from language we must conclude that there are O-O like models in the brain because the semantics is O-O. There might be millions of classes and relationships. And surely every day or night, the brain refactores parts of its model. The roadmap to AGI will probably be top-down and not bottom-up. The bottom-up approach is used by biological evolution. Creating AGI by software engineering means that we first must know where we want to go and then how to go there. Human language and conscious thoughts suggests that AGI must be able to represent the world O-O like at the top-level. So this ability is the answer for the question where we want to go. Again, this does not mean that we must find all the classes and objects. But we must find an algorithm that generates O-O like models of its environment based on its perceptions and some bias where the need for the bias can be proven from reasons of performance. We can expect that the top-level architecture of AGI is the easiest part in an AGI project, because the contents of our own consciousness gives us some hints (but not all) how our own world representation works at the top-level. And this is O-O in my opinion. There is also a phenomenon of associations between patterns (classes). But this is just a question of retrieving information and attention to relevant parts of the O-O model and is no contradiction to the existence of the O-O paradigm. When we go to lower levels, it is clear that difficulties arise. The reason is that we have no possibility for conscious introspection of the low levels in our brain. Science gives us hints mainly for the lowest levels (chemistry, physics...). So the medium layers of AGI will be the most difficult layers. By the way this is also often the case in normal software. In the medium layers there will be base functionalities and the framework for the top-level. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: AW: [agi] How general can be and should be AGI?
Dr. Matthias Heger wrote: Performance not an unimportant question. I assume that AGI has necessarily has costs which grow exponentially with the number of states and actions so that AGI will always be interesting only for toy domains. My assumption is that human intelligence is not truly general intelligence and therefore cannot hold as a proof of existence that AGI is possible. Perhaps we see more intelligence than there really is. Perhaps the human intelligence is to some extend overestimated and an illusion as the free will. Why? In truly general domains every experience of an agent only can be used for the single certain state and action when the experience was made. Every time when your algorithm makes generalizations from known state-action pairs to unknown state-action pairs then this is in fact usage of knowledge about the underlying state-action space or it is just guessing and only a matter of luck. So truly general AGI algorithms must visit every state-action pair at least once to learn what to do in what state. Even in small real world domains the state spaces are so big that it would take longer than the age of the universe to go through all states. For this reason true AGI is impossible and human intelligence must be narrow to a certain degree. I would assert a few things that appear to contradict your assumptions (and a few that suppport them). 1) AGIs will reach conclusions that are not guaranteed to be correct. This allows somewhat lossy compression of the input data. 2) AGIs can exist, but will operate in modes. In AGI mode they will be very expensive and slow. And still be error prone. 3) Humans do have an AGI mode. Probably more than one of them. But it's so expensive to use and so slow that they strive diligently to avoid using it, preferring to rely on simple situation-based models (and discarding most of the input data while doing so). 4) When humans are operating in AGI mode, they are not considering or using ANY real-time data (except to hold and replay notes). The process is too slow. The two AGI modes that I believe people use are 1) mathematics and 2) experiment. Note that both operate in restricted domains, but within those domains they *are* general. (E.g., mathematics cannot generate it's own axioms, postulates, and rules of inference, but given them it is general.) Because of the restricted domains, many problems can't even be addressed by either of them, so I suspect the presence of other AGI modes. Possibly even slower and more expensive to use. I suppose that one could quibble that since the modes I have identified are restricted to particular domains, that they aren't *general* intelligence modes, but as far as I can tell ALL modes of human thought only operate within restricted domains. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: AW: [agi] How general can be and should be AGI?
Charles: as far as I can tell ALL modes of human thought only operate within restricted domains. I literally can't conceive where you got this idea from :). Writing an essay - about, say, the French Revolution, future of AGI, flaws in Hamlet, what you did in the zoo, or any of the other many subject areas of the curriculum - which accounts for, at a very rough estimate, some 50% of problemsolving within education, operates within *which* restricted domain? (And how *did* you arrive at the above idea?) --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: AW: [agi] How general can be and should be AGI?
Mike Tintner wrote: Charles: as far as I can tell ALL modes of human thought only operate within restricted domains. I literally can't conceive where you got this idea from :). Writing an essay - about, say, the French Revolution, future of AGI, flaws in Hamlet, what you did in the zoo, or any of the other many subject areas of the curriculum - which accounts for, at a very rough estimate, some 50% of problemsolving within education, operates within *which* restricted domain? (And how *did* you arrive at the above idea?) Yes, I think of those as being handled largely by specialized, non-general, mechanisms. I suppose that to an extent you could say that it's done via pattern matching, and to that extent it falls under the same model that I've called experimentation. Mainly, though, that's done with specialized language manipulation routines. (I'm not asserting that they are hard-wired. They were built up via lots of time and effort put in via both experimentation and mathematics [in which I include modeling and statistical prediction]). Mathematics and experimentation are extremely broad brushes. That's a part of why they are so slow. French revolution: Learning your history from a teacher or a text isn't a general pattern. It's a short-cut that usually works pretty well. Now if you were talking about going on the ground and doing personal research...then it might count as general intelligence under the category of experimentation. (Note that both mathematics and experimentation are generally necessary to creat new knowledge, rather that copying knowledge from some source that has previously acquired and processed it.) Future of AGI: Creating the future of AGI does, indeed, involve general intelligence. If you follow this list you'll note that it involves BOTH mathematics and experimentation. Flaws in Hamlet: I don't think of this as involving general intelligence. Specialized intelligence, yes, but if you see general intelligence at work there you'll need to be more explicit for me to understand what you mean. Now determining whether a particular deviation from iambic pentameter was a flaw would require a deep human intelligence, but I don't feel that understanding of how human emotions are structured is a part of general intelligence except on a very strongly superhuman level. The level where the AI's theory of your mind was on a par with, or better than, your own. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
AW: AW: [agi] How general can be and should be AGI?
Charles D Hixson [mailto:[EMAIL PROTECTED] The two AGI modes that I believe people use are 1) mathematics and 2) experiment. Note that both operate in restricted domains, but within those domains they *are* general. (E.g., mathematics cannot generate it's own axioms, postulates, and rules of inference, but given them it is general.) Because of the restricted domains, many problems can't even be addressed by either of them, so I suspect the presence of other AGI modes. Possibly even slower and more expensive to use. I suppose that one could quibble that since the modes I have identified are restricted to particular domains, that they aren't *general* intelligence modes, but as far as I can tell ALL modes of human thought only operate within restricted domains. AGI which only operates in restricted domains is no AGI as I understand it But it seems to be, that I use this term in a much stronger sense than most other people. I assume that most people understand AGI as human-like intelligence which refers especially to the repertoire of tasks which can be solved. As I said, 'true AGI' does not use any bias. But any powerful intelligence must use bias because real world state spaces are too complex for 'true AGI'. So AGI as it is used commonly is only approximated and limited AGI but of course much broader than AI of the present and the past. I follow some other people that human-like AGI will probably be build by several narrow AI algorithms that work together. Humans uses and create object oriented descriptions of the world similar to the paradigms of object oriented programming Languages. This paradigm is very powerful in many domains because the inner structure of these domains are in fact object-oriented. But this does not hold in all domains. Among other advantages the object oriented paradigm helps humans to make useful generalizations. For example: A television is a electric appliance. Electric appliances need electric energy. So if there will be some new electric appliance in the future I already know that it will only work if it gets electric energy. On the other hand, the object oriented paradigm is poor for recognition of sounds and voices. We can hardly describe the voice of a person as a set of classes and objects which have some properties, behavior and interact with each other. So the object-oriented paradigm is an example for a very general paradigm but which is not 100% useful in all domains. And the brain has probably not a general monolithic algorithm which finds regularities at all levels. The recognition of regularities in sounds is surely not solved by the same algorithm which learns that houses have windows. The brain even changes its architecture to some degree during lifetime. A baby brain has far more synapses than an adult brain. I think during the first years humans extend their bias which they have from your genes. When they are older, humans can solve many problems but they rely on the bias they obtained during childhood and from their genes. By the way, there is a nice analogy between the brain and the universe: You can't explore your past processes of your own mind of the very first day of your life because your brain changed its inner structure in the first years too much and the algorithm for object oriented patterns probably did not yet exist at that time. The whole universe also will not be able explore its past processes of the big bang. At that time the inner structure changed. There were no atoms, light was scattered always and everywhere. Therefore, we and any possible machine of the universe can only see events some 10 years after the big bang when there were already atoms. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: AW: AW: [agi] How general can be and should be AGI?
--- Dr. Matthias Heger [EMAIL PROTECTED] wrote: Humans uses and create object oriented descriptions of the world similar to the paradigms of object oriented programming Object oriented programming is good for organizing software but I don't think for organizing human knowledge. It is a very rough approximation. We have used O-O for designing ontologies and expert systems (IS-A links, etc), but this approach does not scale well and does not allow for incremental learning from examples. It totally does not work for language modeling, which is the first problem that AI must solve. -- Matt Mahoney, [EMAIL PROTECTED] --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: AW: [agi] How general can be and should be AGI?
Charles: Flaws in Hamlet: I don't think of this as involving general intelligence. Specialized intelligence, yes, but if you see general intelligence at work there you'll need to be more explicit for me to understand what you mean. Now determining whether a particular deviation from iambic pentameter was a flaw would require a deep human intelligence, but I don't feel that understanding of how human emotions are structured is a part of general intelligence except on a very strongly superhuman level. The level where the AI's theory of your mind was on a par with, or better than, your own. Charles, My flabber is so ghasted, I don't quite know what to say. Sorry, I've never come across any remarks quite so divorced from psychological reality. There are millions of essays out there on Hamlet, each one of them different. Why don't you look at a few?: http://www.123helpme.com/search.asp?text=hamlet There are also probably many thousands of critical essays and books on Hamlet, (and may well have been a million written since SHakespeare's time).. The reason people are able to write so many essays is that when you have to write on say whether Hamlet is a tragically flawed hero, you can choose to approach this play (and indeed virtually every other play) from many different angles and domains - tragic theory, psychological - Oedipal complex/ youthful identity crisis, political - young man caught up in corrupt state, moral, Elizabethan dilemma of horror of regicide vs loathing of tyranny, conflict between Hamlet the intellectual and the man of action, inferiority complex in relation to Fortinbras, sexist deconstruction of the suppression of Ophelia, use of poetic imagery and metaphor, stifling self-awareness, and on and on and on The reason over 70 per cent of students procrastinate when writing essays like this about Hamlet, (and the other 20 odd per cent also procrastinate but don't tell the surveys), is in part that it is difficult to know which of the many available approaches to take, and which of the odd thousand lines of text to use as support, and which of innumerable critics to read. And people don't have a neat structure for essay-writing to follow. (And people are inevitably and correctly afraid that it will all take if not forever then far, far too long). This is also the reason why a major percentage of students have difficulty writing an ordered essay and presenting a coherent argument - their essays tend to be cluttered with too many different themes, and keep going off at tangents. In short, essay writing is an excellent example of an AGI in action - a mind freely crossing different domains to approach a given subject from many fundamentally different angles. (If any subject tends towards narrow AI, it is normal as opposed to creative maths). Essay writing also epitomises the NORMAL operation of the human mind. When was the last time you tried to - or succeeded in concentrating for any length of time? As William James wrote of the normal stream of consciousness: Instead of thoughts of concrete things patiently following one another in a beaten track of habitual suggestion, we have the most abrupt cross-cuts and transitions from one idea to another, the most rarefied abstractions and discriminations, the most unheard-of combinations of elements, the subtlest associations of analogy; in a word, we seem suddenly introduced into a seething caldron of ideas, where everything is fizzling and bobbing about in a state of bewildering activity, where partnerships can be joined or loosened in an instant, treadmill routine is unknown, and the unexpected seems the only law. Ditto: The normal condition of the mind is one of informational disorder: random thoughts chase one another instead of lining up in logical causal sequences. Mihaly Csikszentmihalyi Ditto the Dhammapada, Hard to control, unstable is the mind, ever in quest of delight, When you have a mechanical mind that can a) write essays or tell stories or hold conversations [which all present the same basic difficulties] and b) has a fraction of the difficulty concentrating that the brain does and therefore c) a fraction of the flexibility in crossing domains, then you might have something that actually is an AGI. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
Mike, I derived a few things from your response - even enjoyed it. One point passed over too quickly was the question of How knowable is the world? I take this to be a rhetorical question meant to suggest that we need all of it to be considered intelligent. This suggestion seems to be echoed in the statement Which brings us to HOW MANY KINDS OF REPRESENTATIONS OF A SUBJECT.. do we need to form a comprehensive representation of the man - If the implication is that we need it all, the bar is too high - unnecessarily high. 1. We are not building God. The AGI does not need to grasp everything there is to Ben. It doesn't have to conquer the stock market or dominate the institutions of man. It need not have full recall of all important historical events. It need not prefer a philosophy or know the relative value of all things. 2. We are building a machine that performs - intelligent behavior. And because it is to have general intelligent behavior, it must grow. The growth will be done by adopting new xyz? (whatever it finds useful...) For example, as Steve Reed increases the conversational ability of a machine, he will be giving more capability to the unit. With this new capability the unit will have more choices. It will be able to function in an environment where intelligence can be developed / harvested and tested. Who says it won't grow from the advice of those it converses with? 3. Confusion is amplified when there is no distinction between what it is to be intelligent and what it is to be super intelligent. Why make it more difficult than it already is? Why ask the fledgling performer to do what is way beyond it's capacity at inception? 4. If the big deal is will an AGI ever use images? We know that they will. If the question is can they have human comprehension of images? It isn't too much of a stretch to say yeah, probably. As humans we have rich comprehension of many things. Then again, I know many people who think a snake is a snake and that's all they need to know. 5. Minimal system is a target. In the sense of minimal system, I view AGI as a narrow problem. What is the essence of intelligence? - what's required to see intelligent behavior? (qualified to include the broader sense of general intelligence, that is including the growth factors.) Mike, are you saying that there is no such thing as a minimal system? 6. The problem of THIS minimal system is that it is complicated. A few techniques and methods won't do - else such a system exhibiting general intelligent behavior would exist and be growing today. My point - there will continue to be *misunderstanding* if intelligence is viewed without distinguishing mature from fledgling. I'm interested in the minimal system. I consider it my good fortune to have a good seat to observe historic events - I appreciate the project, this list, and it's contributors. Mike Tintner wrote: Matthias: a state description could be: ..I am in a kitchen. The door is open. It has two windows. There is a sink. And three cupboards. Two chairs. A fly is on the right window. The sun is shining. The color of the chair is... etc. etc. .. I think studying the limitations of human intelligence or better to say the predefined innate knowledge in our algorithms is essential to create that what you call AGI. Because only with this knowledge you can avoid the problem of huge state spaces. You did something v. interesting, which is you started to ground the discussion about general intelligence. These discussions are normally almost totally ungrounded as to the SUBJECTS/OBJECTS of intelligence. Essentially, the underlying perspectives of discussions of GI in this field are computational and mathematical re the MEDIUM of intelligence. People basically think along the lines of: how much information can a computer hold, and how can it manipulate that information? But that - the equivalent would be something like: how much can a human brain hold and manipulate? - is not all there is to intelligence. What is totally missing is a philosophical and semiotic perspective. A philosopher looks at things v. differently and asks essentially : how much information can we get about a given subject (and the world generally)? A semioticist asks: how much and what kinds of information about any given subject (or the world generally) can different forms of representation give us? (A verbal description, photo, movie, statue will all give us different forms of info and show different dimensions of a subject). The AI-er asks how much information (about the world) can I and my machine handle? The philosopher: how much information about the world can we actually *get*? - How knowable is the world? ANd what do we have to do to get and present knowledge about the world? If you are
Re: [agi] How general can be and should be AGI?
Stan I'm putting together a detailed paper on this, so overall it will be best to wait for that. My posts today give the barest beginning to my thinking, which is that you start to understand the semiotic requirements for a general intelligence by thinking about the *things* that it must know about, and then look at the dimensions of things that different sign systems - maths/logic/language/schemas/ still images/ dynamic images - *allow* you to see. AGI-ers and indeed most of our culture still think pre-semiotically, and aren't aware that every sign system we use is like a different set of spectacles, and focusses on certain dimensions and problems of things, but totally excludes others. My focus is not so much on the different stages of general intelligence - an evolutionary perspective - although I do think about that. Ironically, it is people who want their AGI's to converse straight away, or variously handle language, who are actually starting in a sense at the godlike, super - human end. It actually takes human intelligence many developmental steps to proceed from being able to process simple, highly specific, concrete, here-and-now this-and-that words for objects and people in immediate scenes, to being able to think in language about vast superclasses of things and creatures spread out over zillions of scenes and billions of years past and future. The idea that you can process, say, the history of the universe is a hard subject to think about, with the same single- or simple-level processing as it's hard to see where the key is in this room is an absurd illusion. Similarly, the idea that you can process all numbers and mathematical entities with the same ease is absurd. It took a long time historically for mathematicians to even dare to think about infinity - a taboo subject until the printing press made it something that could be in part concretely imagined. I'll try then to put out a paper on the semiotics - the bare minimum requirements in terms of sign systems - that I think essential to solve the main problems of AGI, and why, shortly. But re your underlying question - I don't know how tough it will all be. My personal preference is that, as s.o. else just suggested, you guys should link up with some of the roboticists - you both seem to need and complement each other in some ways. But however tough it is, one thing's for sure - it won't do any good to pretend it's easier. AGI will just keep banging its head into brick walls, like it has done for over 50 years. The shorter your cuts, the longer it will actually take. - Original Message - From: Stan Nilsen [EMAIL PROTECTED] To: agi@v2.listbox.com Sent: Tuesday, April 29, 2008 6:52 PM Subject: Re: [agi] How general can be and should be AGI? Mike, I derived a few things from your response - even enjoyed it. One point passed over too quickly was the question of How knowable is the world? I take this to be a rhetorical question meant to suggest that we need all of it to be considered intelligent. This suggestion seems to be echoed in the statement Which brings us to HOW MANY KINDS OF REPRESENTATIONS OF A SUBJECT.. do we need to form a comprehensive representation of the man - If the implication is that we need it all, the bar is too high - unnecessarily high. 1. We are not building God. The AGI does not need to grasp everything there is to Ben. It doesn't have to conquer the stock market or dominate the institutions of man. It need not have full recall of all important historical events. It need not prefer a philosophy or know the relative value of all things. 2. We are building a machine that performs - intelligent behavior. And because it is to have general intelligent behavior, it must grow. The growth will be done by adopting new xyz? (whatever it finds useful...) For example, as Steve Reed increases the conversational ability of a machine, he will be giving more capability to the unit. With this new capability the unit will have more choices. It will be able to function in an environment where intelligence can be developed / harvested and tested. Who says it won't grow from the advice of those it converses with? 3. Confusion is amplified when there is no distinction between what it is to be intelligent and what it is to be super intelligent. Why make it more difficult than it already is? Why ask the fledgling performer to do what is way beyond it's capacity at inception? 4. If the big deal is will an AGI ever use images? We know that they will. If the question is can they have human comprehension of images? It isn't too much of a stretch to say yeah, probably. As humans we have rich comprehension of many things. Then again, I know many people who think a snake is a snake and that's all they need to know. 5. Minimal system is a target. In the sense of minimal system, I view AGI as a narrow problem. What is the essence
AW: [agi] How general can be and should be AGI?
Ben Goertzel [mailto:[EMAIL PROTECTED] wrote 26. April 2008 19:54 Yes, truly general AI is only possible in the case of infinite processing power, which is likely not physically realizable. How much generality can be achieved with how much Processing power, is not yet known -- math hasn't advanced that far yet. My point is not only that 'general intelligence without any limits' would need infinite resources of time and memory. This is trivial of course. What I wanted to say is that any intelligence has to be narrow in a sense if it wants be powerful and useful. There must always be strong assumptions of the world deep in any algorithm of useful intelligence. Let me explain this point in more detail: By useful and powerful intelligence I mean algorithms that do not need resources which grow exponentially with state and action space. Let's take the credit assignment problem of reinforcement learning. The agent has several sensor inputs which builds the perceived state pace of its environment. So if the algorithm is truly general the state space grows exponentially with the number of sensor inputs and the number of time steps it considers of the past. Every pixel of the eyes retina is a part of the state description if you are truly general. And every tiny detail of the past may be important if you are truly general. And even if you are less general and describe your environment not by pixels but by words of common language the state space is huge. For example, a state description could be: ...I am in a kitchen. The door is open. It has two windows. There is a sink. And three cupboards. Two chairs. A fly is on the right window. The sun is shining. The color of the chair is... etc. etc. ... Even this far less general state description would fill pages. So an AGI agent acts in huge state spaces and huge action spaces. It has always to solve the credit assignment problem: Which action in which state is responsible for the current outcome in the current situation. And which action in which state will give me the best outcome? A truly general AI algorithm without much predefined domain-knowledge and suitable for arbitrary state spaces will have to explore the complete state-action space which as I said grows exponentially with sensor inputs and time. I think, every useful intelligence algorithm must always avoid the pitfall of exponential costs and the only way to do this is to be less general and to give the agent more predefined domain knowledge (implicit or explicit, symbolic or non-symbolic, procedural or non-procedural ) Even if you say Human level AI is able to generate its own state spaces. Then there is still the problem that the initial sensory state space is of exponentially extend. So in every useful AGI algorithm there must be certain strong limits as explicit or implicit rules how to represent the world initially and/or how to generalize and build a world representation from experiences. This means, that the only way to avoid the problem of exponentially growth is to hard code implicit or explicit assumptions of the world. And these initial assumptions are the most important limits of any useful intelligence. They are much more important than the restrictions of time and memory. Because with these limits it will probably not be true anymore that you can learn everything and solve any solvable problem if you only get enough resources. The algorithm in itself must has fixed inner limits to be something useful in real world domains. These limits cannot be overcome with experience. Even an algorithm that guesses new algorithms and replaces itself if it can prove that it has found something more useful than itself has fixed statements that it cannot overcome. More important: If you want to make such an algorithm practically useful you have to give it predefined rules how to reduce the huge space of possible algorithms. And again these rules are the more important problem than the lack of memory and space. One could argue that the algorithm can change these rules by own experience. But you can only prove that changing the rules algorithmically enhances the performance if the agent makes good experiences with the new rules. You cannot prove that certain algorithms would not improve your performance if you don't know the algorithms at all. Remember: The rules do not define a certain state or algorithm but they define a reduction of the whole algorithm space the agent can consider while trying to become more powerful. The rules within the algorithm contain knowledge of that what the learning agent does not know itself and cannot learn. Even if you can learn to learn. And learn to learn to learn. And ... Every recursive procedure has to have a non-reducible base and it is clear, that the overall performance and abilities depend crucially on that basic non-reducible procedure. If this procedure is too general, the performance slows exponentially with the space with which this
Re: [agi] How general can be and should be AGI?
On Sun, Apr 27, 2008 at 3:54 AM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: What I wanted to say is that any intelligence has to be narrow in a sense if it wants be powerful and useful. There must always be strong assumptions of the world deep in any algorithm of useful intelligence. From http://nars.wang.googlepages.com/wang-goertzel.AGI_06.pdf Page 5: --- 3.3. General-purpose systems are not as good as special-purpose ones Compared to the previous one, a weaker objection to AGI is to insist that even though general-purpose systems can be built, they will not work as well as special-purpose systems, in terms of performance, efficiency, etc. We actually agree with this judgment to a certain degree, though we do not take it as a valid argument against the need to develop AGI. For any given problem, a solution especially developed for it almost always works better than a general solution that covers multiple types of problem. However, we are not promoting AGI as a technique that will replace all existing domain-specific AI techniques. Instead, AGI is needed in situations where ready-made solutions are not available, due to the dynamic nature of the environment or the insufficiency of knowledge about the problem. In these situations, what we expect from an AGI system are not optimal solutions (which cannot be guaranteed), but flexibility, creatively, and robustness, which are directly related to the generality of the design. In this sense, AGI is not proposed as a competing tool to any AI tool developed before, by providing better results, but as a tool that can be used when no other tool can, because the problem is unknown in advance. --- Pei --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
AW: [agi] How general can be and should be AGI?
Performance not an unimportant question. I assume that AGI has necessarily has costs which grow exponentially with the number of states and actions so that AGI will always be interesting only for toy domains. My assumption is that human intelligence is not truly general intelligence and therefore cannot hold as a proof of existence that AGI is possible. Perhaps we see more intelligence than there really is. Perhaps the human intelligence is to some extend overestimated and an illusion as the free will. Why? In truly general domains every experience of an agent only can be used for the single certain state and action when the experience was made. Every time when your algorithm makes generalizations from known state-action pairs to unknown state-action pairs then this is in fact usage of knowledge about the underlying state-action space or it is just guessing and only a matter of luck. So truly general AGI algorithms must visit every state-action pair at least once to learn what to do in what state. Even in small real world domains the state spaces are so big that it would take longer than the age of the universe to go through all states. For this reason true AGI is impossible and human intelligence must be narrow to a certain degree. -Ursprüngliche Nachricht- Von: Pei Wang [mailto:[EMAIL PROTECTED] Gesendet: Sonntag, 27. April 2008 13:50 An: agi@v2.listbox.com Betreff: Re: [agi] How general can be and should be AGI? On Sun, Apr 27, 2008 at 3:54 AM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: What I wanted to say is that any intelligence has to be narrow in a sense if it wants be powerful and useful. There must always be strong assumptions of the world deep in any algorithm of useful intelligence. From http://nars.wang.googlepages.com/wang-goertzel.AGI_06.pdf Page 5: --- 3.3. General-purpose systems are not as good as special-purpose ones Compared to the previous one, a weaker objection to AGI is to insist that even though general-purpose systems can be built, they will not work as well as special-purpose systems, in terms of performance, efficiency, etc. We actually agree with this judgment to a certain degree, though we do not take it as a valid argument against the need to develop AGI. For any given problem, a solution especially developed for it almost always works better than a general solution that covers multiple types of problem. However, we are not promoting AGI as a technique that will replace all existing domain-specific AI techniques. Instead, AGI is needed in situations where ready-made solutions are not available, due to the dynamic nature of the environment or the insufficiency of knowledge about the problem. In these situations, what we expect from an AGI system are not optimal solutions (which cannot be guaranteed), but flexibility, creatively, and robustness, which are directly related to the generality of the design. In this sense, AGI is not proposed as a competing tool to any AI tool developed before, by providing better results, but as a tool that can be used when no other tool can, because the problem is unknown in advance. --- Pei --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
If by truly general you mean absolutely general, I agree it is not possible, but it is not what we are after. Again, I hope you find out what people are doing under the name AGI, then make your argument against it, rather than against the AGI in your imagination. For example, I fully agree that visiting every state-action pair is hopeless, but who in AGI is doing that or suggesting that? Just because traditional AI is trapped by this methodology doesn't mean there is no other possibility. Who said AI systems must do state-based planning? I'm not trying to convince you that AGI can be achieved --- that is what people are exploring --- but that you should not assume the traditional AI has tried all possibilities, and there cannot be anything new. Of course every intelligent system (human or computer) has its limit (nobody denied that), but that limit is fundamentally different from the limit of the current AI systems. Pei On Sun, Apr 27, 2008 at 8:46 AM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: Performance not an unimportant question. I assume that AGI has necessarily has costs which grow exponentially with the number of states and actions so that AGI will always be interesting only for toy domains. My assumption is that human intelligence is not truly general intelligence and therefore cannot hold as a proof of existence that AGI is possible. Perhaps we see more intelligence than there really is. Perhaps the human intelligence is to some extend overestimated and an illusion as the free will. Why? In truly general domains every experience of an agent only can be used for the single certain state and action when the experience was made. Every time when your algorithm makes generalizations from known state-action pairs to unknown state-action pairs then this is in fact usage of knowledge about the underlying state-action space or it is just guessing and only a matter of luck. So truly general AGI algorithms must visit every state-action pair at least once to learn what to do in what state. Even in small real world domains the state spaces are so big that it would take longer than the age of the universe to go through all states. For this reason true AGI is impossible and human intelligence must be narrow to a certain degree. -Ursprüngliche Nachricht- Von: Pei Wang [mailto:[EMAIL PROTECTED] Gesendet: Sonntag, 27. April 2008 13:50 An: agi@v2.listbox.com Betreff: Re: [agi] How general can be and should be AGI? On Sun, Apr 27, 2008 at 3:54 AM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: What I wanted to say is that any intelligence has to be narrow in a sense if it wants be powerful and useful. There must always be strong assumptions of the world deep in any algorithm of useful intelligence. From http://nars.wang.googlepages.com/wang-goertzel.AGI_06.pdf Page 5: --- 3.3. General-purpose systems are not as good as special-purpose ones Compared to the previous one, a weaker objection to AGI is to insist that even though general-purpose systems can be built, they will not work as well as special-purpose systems, in terms of performance, efficiency, etc. We actually agree with this judgment to a certain degree, though we do not take it as a valid argument against the need to develop AGI. For any given problem, a solution especially developed for it almost always works better than a general solution that covers multiple types of problem. However, we are not promoting AGI as a technique that will replace all existing domain-specific AI techniques. Instead, AGI is needed in situations where ready-made solutions are not available, due to the dynamic nature of the environment or the insufficiency of knowledge about the problem. In these situations, what we expect from an AGI system are not optimal solutions (which cannot be guaranteed), but flexibility, creatively, and robustness, which are directly related to the generality of the design. In this sense, AGI is not proposed as a competing tool to any AI tool developed before, by providing better results, but as a tool that can be used when no other tool can, because the problem is unknown in advance. --- Pei --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303
AW: [agi] How general can be and should be AGI?
Ok. Maybe we mean different things under the name AGI. I agree that traditional AI is just the beginning. And even if human intelligence is no proof for that what I mean with AGI it is clear that human intelligence is far way more powerful than any AI until now. But perhaps only for subtle reasons or reasons we will not know in 100 years, who know. I am convinced that super human intelligence is possible. But mainly because we will use faster and more hardware than our brain. Biology has created intelligence by trial and error and used billions of years and trillions of animals for this. The goal to do it better within 10 or 20 years from now and from scratch seems to me way too ambitious. I think studying the limitations of human intelligence or better to say the predefined innate knowledge in our algorithms is essential to create that what you call AGI. Because only with this knowledge you can avoid the problem of huge state spaces. -Ursprüngliche Nachricht- Von: Pei Wang [mailto:[EMAIL PROTECTED] Gesendet: Sonntag, 27. April 2008 15:03 An: agi@v2.listbox.com Betreff: Re: [agi] How general can be and should be AGI? If by truly general you mean absolutely general, I agree it is not possible, but it is not what we are after. Again, I hope you find out what people are doing under the name AGI, then make your argument against it, rather than against the AGI in your imagination. For example, I fully agree that visiting every state-action pair is hopeless, but who in AGI is doing that or suggesting that? Just because traditional AI is trapped by this methodology doesn't mean there is no other possibility. Who said AI systems must do state-based planning? I'm not trying to convince you that AGI can be achieved --- that is what people are exploring --- but that you should not assume the traditional AI has tried all possibilities, and there cannot be anything new. Of course every intelligent system (human or computer) has its limit (nobody denied that), but that limit is fundamentally different from the limit of the current AI systems. Pei On Sun, Apr 27, 2008 at 8:46 AM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: Performance not an unimportant question. I assume that AGI has necessarily has costs which grow exponentially with the number of states and actions so that AGI will always be interesting only for toy domains. My assumption is that human intelligence is not truly general intelligence and therefore cannot hold as a proof of existence that AGI is possible. Perhaps we see more intelligence than there really is. Perhaps the human intelligence is to some extend overestimated and an illusion as the free will. Why? In truly general domains every experience of an agent only can be used for the single certain state and action when the experience was made. Every time when your algorithm makes generalizations from known state-action pairs to unknown state-action pairs then this is in fact usage of knowledge about the underlying state-action space or it is just guessing and only a matter of luck. So truly general AGI algorithms must visit every state-action pair at least once to learn what to do in what state. Even in small real world domains the state spaces are so big that it would take longer than the age of the universe to go through all states. For this reason true AGI is impossible and human intelligence must be narrow to a certain degree. -Ursprüngliche Nachricht- Von: Pei Wang [mailto:[EMAIL PROTECTED] Gesendet: Sonntag, 27. April 2008 13:50 An: agi@v2.listbox.com Betreff: Re: [agi] How general can be and should be AGI? On Sun, Apr 27, 2008 at 3:54 AM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: What I wanted to say is that any intelligence has to be narrow in a sense if it wants be powerful and useful. There must always be strong assumptions of the world deep in any algorithm of useful intelligence. From http://nars.wang.googlepages.com/wang-goertzel.AGI_06.pdf Page 5: --- 3.3. General-purpose systems are not as good as special-purpose ones Compared to the previous one, a weaker objection to AGI is to insist that even though general-purpose systems can be built, they will not work as well as special-purpose systems, in terms of performance, efficiency, etc. We actually agree with this judgment to a certain degree, though we do not take it as a valid argument against the need to develop AGI. For any given problem, a solution especially developed for it almost always works better than a general solution that covers multiple types of problem. However, we are not promoting AGI as a technique that will replace all existing domain-specific AI techniques. Instead, AGI is needed in situations where ready-made solutions are not available, due to the dynamic nature of the environment or the insufficiency of knowledge about the problem. In these situations, what we
Re: [agi] How general can be and should be AGI?
On Sun, Apr 27, 2008 at 3:54 AM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: Ben Goertzel [mailto:[EMAIL PROTECTED] wrote 26. April 2008 19:54 Yes, truly general AI is only possible in the case of infinite processing power, which is likely not physically realizable. How much generality can be achieved with how much Processing power, is not yet known -- math hasn't advanced that far yet. My point is not only that 'general intelligence without any limits' would need infinite resources of time and memory. This is trivial of course. What I wanted to say is that any intelligence has to be narrow in a sense if it wants be powerful and useful. There must always be strong assumptions of the world deep in any algorithm of useful intelligence. This is a consequence of the No Free Lunch theorem, essentially, isn't it? http://en.wikipedia.org/wiki/No_free_lunch_in_search_and_optimization With infinite resources you use exhaustive search (like AIXI or the Godel Machine) ... with finite resources you can't afford it, so you need to use (explicitly or implicitly) search that is guided by some inductive biases. See Eric Baum's book What Is Thought? for much discussion on genetically encoded inductive bias and its role in AI. -- Ben G --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
Matthias: a state description could be: ...I am in a kitchen. The door is open. It has two windows. There is a sink. And three cupboards. Two chairs. A fly is on the right window. The sun is shining. The color of the chair is... etc. etc. .. I think studying the limitations of human intelligence or better to say the predefined innate knowledge in our algorithms is essential to create that what you call AGI. Because only with this knowledge you can avoid the problem of huge state spaces. You did something v. interesting, which is you started to ground the discussion about general intelligence. These discussions are normally almost totally ungrounded as to the SUBJECTS/OBJECTS of intelligence. Essentially, the underlying perspectives of discussions of GI in this field are computational and mathematical re the MEDIUM of intelligence. People basically think along the lines of: how much information can a computer hold, and how can it manipulate that information? But that - the equivalent would be something like: how much can a human brain hold and manipulate? - is not all there is to intelligence. What is totally missing is a philosophical and semiotic perspective. A philosopher looks at things v. differently and asks essentially : how much information can we get about a given subject (and the world generally)? A semioticist asks: how much and what kinds of information about any given subject (or the world generally) can different forms of representation give us? (A verbal description, photo, movie, statue will all give us different forms of info and show different dimensions of a subject). The AI-er asks how much information (about the world) can I and my machine handle? The philosopher: how much information about the world can we actually *get*? - How knowable is the world? ANd what do we have to do to get and present knowledge about the world? If you are truly serious here, I suggest, you have to look at intelligence from both perspectives. You took a kitchen as a possible subject to ground the discussion. Why not take something easier to think about - to consider the difficulties of getting to know the world - a human being. Take one at random: http://lifeboat.com/board/ben.goertzel.jpg What does anyone, any society or any intelligence need to be a) intelligent - to - ultimately b) omniscient about this man? How many disciplines of knowledge studying how many LEVELS OF THE SUBJECT - levels of this man and his body, behaviour and relationships do we need to bring in? Presumably we need somewhere between something and everything our culture has to offer - every branch of science - psychology, social psychology, biopsychology, social anthropology, behavioural economics, cognitive science, neuroscience, down to cardiology, gastroenterology, immunology ... down to biochemistry, molecular biology, genetics - focussing on every part of his behaviour, and every part or subsystem of his body. (Would you want a total systems science view which would attempt to integrate all their views into one totally inegrated model of the man? Our culture doesn't offer such a thing only a piecemeal view, but maybe you'd like to attempt one?) And those are just the generalists. Then we really ought to bring in somewhere between some and all kinds of the arts - they specialise in individual portraits. Novelists, painters, sculptors, moviemakers, cartoonists etc. A Scorsese at least to do justice to his titanic struggles. They can all show us different of dimensions of this man. Which brings us to HOW MANY KINDS OF REPRESENTATIONS OF A SUBJECT.. do we need to form a comprehensive representation of the man - textual, references on Google, mathematicial, photographic, drawing, cartoon, movies, statues, 3-d molecular models, holograms, tax returns, bank statements how many scientific representations - mammogram, cardiogram, urine samples, skin samples, biopsies, blood tests... And then how much PERSONAL INTERACTION WITH THE SUBJECT is needed. Should you have interviewed, worked with him, partied with him, had sex with im? - And the SUBJECT'S RELATIONS ... should you know his family, friends etc.? How extensive REPRESENTATIONS OF THE SUBJECT'S ENVIRONMENT... his home, office, car, beat-up chair, clothes etc...local neighbourhood, town, etc.. And what DEGREE OF EMBODIMENT should you, the knower, - or your computer - have? Because, obviously, you can only identify with any given subject to the extent that you have a similar/the same body. Hence philosophy's what's it like to be a bat? and how can you know *my* qualia? obsessions. Even God, according to some religions, had to become flesh to know humans. Ultimately, I suggest, PERFECTION ...near godlike knowledge and intelligence would involve having a PERFECT REPLICA OF THE SUBJECT AND HIS ENVIRONMENT...
Re: [agi] How general can be and should be AGI?
2008/4/27 Dr. Matthias Heger [EMAIL PROTECTED]: Ben Goertzel [mailto:[EMAIL PROTECTED] wrote 26. April 2008 19:54 Yes, truly general AI is only possible in the case of infinite processing power, which is likely not physically realizable. How much generality can be achieved with how much Processing power, is not yet known -- math hasn't advanced that far yet. My point is not only that 'general intelligence without any limits' would need infinite resources of time and memory. This is trivial of course. What I wanted to say is that any intelligence has to be narrow in a sense if it wants be powerful and useful. There must always be strong assumptions of the world deep in any algorithm of useful intelligence. I am probably the one on this list the closest to the position you think AGI means. I would agree. Any algorithms needs to be very specific to be useful. However the *architecture*, of an AGI needs to be general (by this I mean capable of instantiating any TM equivalent function, from input and current state to output and current state). So I think the the lowest level of the system space should be massive as you argue against. However I would not make it a search space, as such, with a fixed method searching it. On its own, it should be passive, however it is be able to have active programs within it. As these are programs on there own they can search the space of possible programs. These programs could search sub spaces of the entire space, or get information from the outside about which subspaces to search. However, there is no limit to which subspaces they do actually search. What makes my approach different to a bog standard computer system, is that it would guide the searching of the programs within it, by acting as reinforcement based ratchet. Those programs with the most reinforcement, that act sensibly, will be able to protect and expand the influence they have over the system. With the right internal programs and environment, this will look as if the system has a goal for what it is trying to become. See this post for more details. http://www.mail-archive.com/agi@v2.listbox.com/msg02892.html Every recursive procedure has to have a non-reducible base and it is clear, that the overall performance and abilities depend crucially on that basic non-reducible procedure. If this procedure is too general, the performance slows exponentially with the space with which this basic procedure works. There are recursive procedures that abandon the base, see for example booting a machine. Will Pearson --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
AW: [agi] How general can be and should be AGI?
Mike Tintner wrote What is totally missing is a philosophical and semiotic perspective. A philosopher looks at things v. differently and asks essentially : how much information can we get about a given subject (and the world generally)? A semioticist asks: how much and what kinds of information about any given subject (or the world generally) can different forms of representation give us? (A verbal description, photo, movie, statue will all give us different forms of info and show different dimensions of a subject). The AI-er asks how much information (about the world) can I and my machine handle? The philosopher: how much information about the world can we actually *get*? - How knowable is the world? ANd what do we have to do to get and present knowledge about the world? I think, the typical AIer asks for clever algorithms and source code. Certain AIer see their problem as a part of control theory. Other see Their problem as the problem of emulating the biological brain. But very few ask about the regularities of our world. May be, because they think that this should be done by the intelligent software we want to build. But I am convinced that goal directed engineering to develop AGI is only possible if we model the most basic regularities of our universe in the AGI algorithms. Humans make experiences in single states and can generalize the new knowledge to huge domains. Most AIer ask only: How do we generalize? But the answer depends on the question: Why is it even possible that we may generalize? And this question is rarely considered. This was the point I wanted to say, that our universe does not only help life to evolve but it seems to be very friendly for intelligence because our universe is full of regularities at all levels from microcosm to macrocosm. And useful intelligence is only possible in a world with regularities because only with regularities you can avoid to search through trillions of states. For example: I can talk tomorrow with a person who I have never seen before. I can do this just from my social experiences of the past with other people. Why is this possible? Another example: I see a mosquito for the first time in my life. I hear the sound. I see how it lands on my right arm. And I see it flying away. Finally my skin becomes red at that place and I feel the little pain at my. Why can I conclude the mosquito is the reason for the pain? Why can I know that the same could happen, if the mosquito would land on my shoulder? Why do I know that the room is not important for the phenomenon of the pain? I think, an AIer must ask such questions. And he has to see it from both perspectives: On the one hand the software engineer who designs the intelligent algorithm. On the other hand the scientist who thinks about nature. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
[agi] How general can be and should be AGI?
How general should be AGI? When I heard the term AGI for the first time, I had to think about the general problem solver from 1959 (http://en.wikipedia.org/wiki/General_Problem_Solver). It solved a few simple problems but was overstrained with real world problems. Second, there is Gödel's theorem which shows that there can not be a complex machine that can can generate both complete and uncontradictory knowledge. There is another theorem that shows that 100% AGI is impossible: Alan Turing proved in 1936 that a general algorithm to solve the halting problem for all possible program-input pairs cannot exist. On the other hand, we think that AGI is possible because we believe that we ourselves ARE AGI systems. But as the theorems show without any doubt: Perfect AGI is impossible. Of course I am convinced that there can be systems, which are far more intelligent than humans. But even these systems will have their limits. Perhaps further research of our own limits can help us to construct more intelligent machines. Perhaps the mechanisms behind human intelligence are so powerful because they are not designed to be a real general problem solver. Intelligence has a lot to do with recognizing regularities in patterns of signals which are obtained from the environment. We can see humans to be very powerful in this ability. But are we real powerful to recogonize GENERAL regularities??? A simple example shows, that this is by far not the case: We can recognize very detailed information from the environment by our eyes. So we can think, that our optical sense is a general pattern recognizer. But imagine you would record a soundwave of a speaking man and visualize the wave on a screen. It would be impossible for you to recognize the words or even the voice of the person. I am sure that even if you would practice a child for years with these patterns it would not learn to understand the voice and sentences. Perhaps a slow and errorful recognition of some words. But by far not so powerful as our acousthesia. This shows that our optical sense is not able to recognize general patterns of our environment. And by the way: The child would not gain conscious phenomenons like qualia when analyzing the sound waves. Our optical intelligence pattern recognizer is NOT AGI. It is narrow AI in this sense. Assumption 1: ### Most powerful intelligence and most general intelligence are not possible at the same time. ### A system which has most general intelligence will suffer from huge problems of complexity. So if we design an architecture which can evolve to very very general intelligence it will be very very probably need too many time and memory so that it can be only of theoretical interest. So one main problem of AGI is to design it general but not too general. And one of the main questions will be which features and domain knowledge should be hard coded and how. If we define intelligence to be the ability to solve complex problems in complex environments we should ask what are adequate limits of complexity. Life can only evolve in environments with very narrow conditions. I think this is similar with intelligence: Assumption 2: ### Intelligence can work and evolve only in environments with limited conditions. ### Nature is very very complex because there are so many particles which interacts with each other. It is important for intelligence, that knowledge of every single particle and fundamental laws of physics is not necessary to make predictions about the environment The change of day and night is an example for a regularity which can be predicted with high accuracy with very low knowledge of details of the environment. In a world with low structure or rapid change of structure and regularities intelligence is for sure not possible or very difficult at least. Our world is a hierarchical world with encapsulated levels. You can see regularities on high levels without the knowledge of the details below this level. This is certainly a key feature of our world without that intelligent life could not evolve at all. So I see the following interesting questions for AGI: What restrictions are adequate or necessary for a practical AGI system to obtain a good compromise of general and powerful intelligence? What are the detailed conditions of the environment, that are necessary for intelligent systems. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
From http://nars.wang.googlepages.com/wang-goertzel.AGI_06.pdf page 5: --- In the current context, when we say that the human mind or an AGI system is general purpose, we do not mean that it can solve all kinds of problems in all kinds of domains, but that it has the potential to solve any problem in any domain, given proper experience. Non-AGI systems lack such a potential. --- That paper also addressed the issue of general potential vs. domain knowledge. I agree with you that an intelligence solving all kinds of problems in all kinds of domains is impossible, though I don't think the conclusions of Gödel and Turing are the major reason (or even that relevant) here. My arguments are in http://nars.wang.googlepages.com/wang.AI_Misconceptions.pdf Pei On Sat, Apr 26, 2008 at 6:35 AM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: How general should be AGI? When I heard the term AGI for the first time, I had to think about the general problem solver from 1959 (http://en.wikipedia.org/wiki/General_Problem_Solver). It solved a few simple problems but was overstrained with real world problems. Second, there is Gödel's theorem which shows that there can not be a complex machine that can can generate both complete and uncontradictory knowledge. There is another theorem that shows that 100% AGI is impossible: Alan Turing proved in 1936 that a general algorithm to solve the halting problem for all possible program-input pairs cannot exist. On the other hand, we think that AGI is possible because we believe that we ourselves ARE AGI systems. But as the theorems show without any doubt: Perfect AGI is impossible. Of course I am convinced that there can be systems, which are far more intelligent than humans. But even these systems will have their limits. Perhaps further research of our own limits can help us to construct more intelligent machines. Perhaps the mechanisms behind human intelligence are so powerful because they are not designed to be a real general problem solver. Intelligence has a lot to do with recognizing regularities in patterns of signals which are obtained from the environment. We can see humans to be very powerful in this ability. But are we real powerful to recogonize GENERAL regularities??? A simple example shows, that this is by far not the case: We can recognize very detailed information from the environment by our eyes. So we can think, that our optical sense is a general pattern recognizer. But imagine you would record a soundwave of a speaking man and visualize the wave on a screen. It would be impossible for you to recognize the words or even the voice of the person. I am sure that even if you would practice a child for years with these patterns it would not learn to understand the voice and sentences. Perhaps a slow and errorful recognition of some words. But by far not so powerful as our acousthesia. This shows that our optical sense is not able to recognize general patterns of our environment. And by the way: The child would not gain conscious phenomenons like qualia when analyzing the sound waves. Our optical intelligence pattern recognizer is NOT AGI. It is narrow AI in this sense. Assumption 1: ### Most powerful intelligence and most general intelligence are not possible at the same time. ### A system which has most general intelligence will suffer from huge problems of complexity. So if we design an architecture which can evolve to very very general intelligence it will be very very probably need too many time and memory so that it can be only of theoretical interest. So one main problem of AGI is to design it general but not too general. And one of the main questions will be which features and domain knowledge should be hard coded and how. If we define intelligence to be the ability to solve complex problems in complex environments we should ask what are adequate limits of complexity. Life can only evolve in environments with very narrow conditions. I think this is similar with intelligence: Assumption 2: ### Intelligence can work and evolve only in environments with limited conditions. ### Nature is very very complex because there are so many particles which interacts with each other. It is important for intelligence, that knowledge of every single particle and fundamental laws of physics is not necessary to make predictions about the environment The change of day and night is an example for a regularity which can be predicted with high accuracy with very low knowledge of details of the environment. In a world with low structure or rapid change of structure and regularities intelligence is for sure not possible or very difficult at least. Our world is a hierarchical world with encapsulated levels. You can see regularities on high levels without the knowledge of the details below this level. This is certainly a key
AW: [agi] How general can be and should be AGI?
In my opinion you can apply Gödel's theorem to prove that 100% AGI is not possible in this world if you apply it not to a hypothetical machine or human being but to the whole universe which can be assumed to be a closed system. The axioms are the laws of physics. Then, everything what happens in this world is the application of these axioms. And every step of this application is without any fault if we suppose that our universe does not change the laws of physics. So, with Goedel, the whole universe cannot generate a set of statements about itself which are both complete and without any contradictions. And therefore a machine which is part of the universe cannot have this ability too. But my main point was not to think about the question whether perfect AGI is possible. I mainly wanted to point out, that we ourselves have strong limits and are in a sense narrow AI systems instead of AGI systems. The example with the visualized sound wave shows that we use very specialized pattern algorithms instead of general ones. And of course biology use it for reasons of performance. Perhaps it is possible for a human to see the patterns in a sound wave if he has enough time. But this would be thousand fold slower than the specialized pattern recognizer for sound signals. This shows that human intelligence is not build from general pattern algorithms in the brain but from algorithms that are specialized for patterns of a specialized environment or at least are tuned for specialized patterns. From this raises the question whether it makes sense to think about pattern algorithms that works with most patterns in this world. This question is mainly a question of performance. And my point was the assumption, that we can buy general intelligence only for hopelessly many costs of time and memory. Another example: Imagine a child who makes the first experience with pain when touching a hot hotplate in a kitchen. The child will learn not to touch the hotplate. But this task is very hard if you want to solve it with a general algorithm with low domain-knowledge. If you feel pain in your hand, what was the reason? If you think in the AGI way it could be the open window in the kitchen, the sandwich you have eaten an hour ago, the fly on the desk, your blue shirt ... and after trillions of other possible reasons the collision of the hand with the plate which is obvious for us but not obvious for any algorithm without domain knowledge. Well you can find the reason with more experience. But how many tries do you need with AGI? Trillions! Because with the AGI approach you can by definition not rule out anything. At least you have to use AGI learning algorithms with massive predefined rules of generalization. So even if we find clever AGI algorithms, the power will mainly depend on tuning it to work in special real world problems. -Ursprüngliche Nachricht- Von: Pei Wang [mailto:[EMAIL PROTECTED] Gesendet: Samstag, 26. April 2008 14:16 An: agi@v2.listbox.com Betreff: Re: [agi] How general can be and should be AGI? From http://nars.wang.googlepages.com/wang-goertzel.AGI_06.pdf page 5: --- In the current context, when we say that the human mind or an AGI system is general purpose, we do not mean that it can solve all kinds of problems in all kinds of domains, but that it has the potential to solve any problem in any domain, given proper experience. Non-AGI systems lack such a potential. --- That paper also addressed the issue of general potential vs. domain knowledge. I agree with you that an intelligence solving all kinds of problems in all kinds of domains is impossible, though I don't think the conclusions of Gödel and Turing are the major reason (or even that relevant) here. My arguments are in http://nars.wang.googlepages.com/wang.AI_Misconceptions.pdf Pei --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
On Sat, Apr 26, 2008 at 2:35 PM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: How general should be AGI? If all you aim for is a system that has unlimited potential, then a Universal Turing Machine is as far as you need to go, and as far as you can go. A more important goal to be build a system that can learn to do relevant things in reasonable time. Our world is a hierarchical world with encapsulated levels. You can see regularities on high levels without the knowledge of the details below this level. This is certainly a key feature of our world without that intelligent life could not evolve at all. I think you captured the bias that AGI system needs to have in this quote quite well. It's probably a good summary of what a system must be able to do to be deemed intelligent in the sense people are intelligent: to be able to simulate the environment on different levels of detail, focusing attention of different parts of the structure, moving the attention to more general levels or more detailed ones, to different parts of the structure or to the future and the past of its development. Starting from this basis, it should be possible to develop specialized subsystems that solve specific problems not amenable to such description. -- Vladimir Nesov [EMAIL PROTECTED] --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
Tell me: what are the algorithms that will force you to process this image in an inevitable way (and what is that way?): http://honolulu.hawaii.edu/distance/sci122/Programs/p3/Rorschach.gif (Oh - and a, linas, Bob, Mark, et al - can we agree that there is no way for maths to process that image, period?) No. I strongly disagree with your assertion. What you believe you are processing (w)holistically can easily be broken down into a series of parts with scaling, translation, transposition, and other standard operators. What a rorschach epitomises is the human ability to see just about anything in anything (and you can't get much more general than that) - a solar system in an atom, a hard, metal computer in a grey, gooey brain, or a penis in a rocket and the odd, million other items ... Again, just all parts processed with normal visual operators. Imagination. That's what's at the heart of adaptivity and general intelligence - and that's what AGI totally - and systemically - lacks. I agree -- but vision is *NOT* the source and not required. You are merely faked out by the fact that it is the primary sense of human beings. Face it. Your arguments are *not* compelling. You seem to believe that whining the same thing over and over again like a petulant child is going to change minds. Trust me, unless you come up with a logical reason, more and more people are just going to stop listening to you. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
AW: [agi] How general can be and should be AGI?
On Samstag, 26. April 2008 17:00 Pei Wang [mailto:[EMAIL PROTECTED] wrote to many people, including me, this is exactly what AGI is after: a baby with all kinds of potentials, not an adult that can do everything. I understand AGI in the same way but even the term all kind of potentials seems to me a wish which is not possible and which also no human baby really has. Let's take the halting problem(http://en.wikipedia.org/wiki/Halting_problem). It is fact that no Turing machine (http://en.wikipedia.org/wiki/Turing_machine) can solve it for all program-input pairs. I think, your argumentation is that an AGI system (e.g. human being) can solve any halting problem because it can change over time by making more and more experiences. But the even the experience making human being can be regarded as a turing machine with a fixed and finite algorithm. All knowledge it can obtain is already implicit there - in the universe. The universe can be modeled as part of the infinite tape of the turing machine. The computer or the brain is the finite table of the turing machine. Every making of experience of AGI can be modeled as reading data from the turing tape. Even if you think of a human being who uses more and more pieces of paper to expand his knowledge and behavior or a computer who grow into space, then simply model the whole universe to be the finite table of the turing machine and the tape at the same time. So even if you point out that AGI should mean to have ALL potentials instead of having ALL abilities, AGI is impossible and can only be approximated. I therefore prefer the term human level AGI My point is that the theoretical thoughts are strong evidence for fundamental limitations of human intelligence and that perhaps (=assumption) these and further limitations are to a certain degree even necessary to solve the problem of the need for huge time, memory, learning steps. Our universe is not only well build for life to evolve. It has also allowed human level AGI to evolve from very narrow forms of intelligence. I like the goal to create AGI but I fear, we want to make it too general and can therefore not overcome problems of complexity. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
On Sat, Apr 26, 2008 at 10:03 AM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: In my opinion you can apply Gödel's theorem to prove that 100% AGI is not possible in this world if you apply it not to a hypothetical machine or human being but to the whole universe which can be assumed to be a closed system. Please consult the works of Marcus Hutter (Universal AI) and Juergen Schmidhuber (Godel Machine). These thoughts are not new. Yes, truly general AI is only possible in the case of infinite processing power, which is likely not physically realizable. How much generality can be achieved with how much processing power, is not yet known -- math hasn't advanced that far yet. Humans are not totally general, yet are much more general than any of the AI systems yet built -- Ben G --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
On Sat, Apr 26, 2008 at 9:39 PM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: I think, your argumentation is that an AGI system (e.g. human being) can solve any halting problem because it can change over time by making more and more experiences. But the even the experience making human being can be regarded as a turing machine with a fixed and finite algorithm. It's pretty bald to expect that many people could've been missing such a flaw in your straw man for years, isn't it? -- Vladimir Nesov [EMAIL PROTECTED] --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
MT: http://honolulu.hawaii.edu/distance/sci122/Programs/p3/Rorschach.gif (Oh - and a, linas, Bob, Mark, et al - can we agree that there is no way for maths to process that image, period?) Mark:No. I strongly disagree with your assertion. What you believe you are processing (w)holistically can easily be broken down into a series of parts with scaling, translation, transposition, and other standard operators. Mark, Let's see if you can actually put forward an idea, as opposed to shouting. You've missed the point. What a human does in looking at a rorschach is to see - i.e. compare it with - a recognizable object or creature - a bat, for instance, or an ant, or a gargoyle. So what you must tell me is how your or any geometrical system of analysis is going to be able to take a rorschach and come up similarly with a recognizable object or creature. Bear in mind, your system will be given no initial clues as to what objects or creatures are suitable as potential comparisons. It can by all means have a large set of visual images in memory, as we do. But you must tell me how your system will connect the rorschach with any of those images, such as a bat, - by *geometrical* means. Of course, a geometrical system can be used to *analyse* an individual rorschach into some set of geometric forms - but only by hand, individually, on a one-off basis - and imperfectly. There is no geometrical *formula* for analysing rorschachs, because they can take an unlimited and non-formulaic variety of shapes, just as there is no geometrical formula for analysing the diverse shapes of living creatures, like bats, ants and human beings. So there is, by extension, no geometrical, or indeed any other formulaic means to *compare* a rorschach with any object or creature. Nor is there any geometrical means to compare *any* irregular objects - a slug and, say, a human being walking along, a purse, say, and a vagina, a rock and a chair, a moustache and a walrus. If you think there is, then you obviously have solved some of the most important, unsolved problems of AGI, such as analogy, metaphor and creativity. You've also turned geometers into designers and artists You also may have gone some way to solving the problem of lookup - the way the brain can find a similar image in just a few steps, where computers that can only search blindly take millions of steps, and may still come up with nothing. So I await your geometric solution to this problem - (a mere statement of principle will do) - with great interest. Well, actually no. Your answer is broadly predictable - you 1) won't have any idea here 2) will have nothing to say to the point and 3) be, as usual, all bark and no bite - all insults and no ideas. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
On Sat, Apr 26, 2008 at 8:09 PM, Mike Tintner wrote: So what you must tell me is how your or any geometrical system of analysis is going to be able to take a rorschach and come up similarly with a recognizable object or creature. Bear in mind, your system will be given no initial clues as to what objects or creatures are suitable as potential comparisons. It can by all means have a large set of visual images in memory, as we do. But you must tell me how your system will connect the rorschach with any of those images, such as a bat, - by *geometrical* means. snip This is called Content-based image retrieval (CBIR), also known as query by image content (QBIC) and content-based visual information retrieval (CBVIR) is the application of computer vision to the image retrieval problem, that is, the problem of searching for digital images in large databases. http://en.wikipedia.org/wiki/CBIR This is a hot area of computer research, with many test systems. (see article). Nothing to do with AGI, of course. Every post from Mike seems to be yet another different way of saying 'You're all wrong!' Are you sure you want to be on this list, Mike? BillK --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
AW: [agi] How general can be and should be AGI?
Don't understand your point fully. Perhaps my English is too bad. I have had the impression, that pei wang thought that gödels theorem and the halting problem do not apply for human beings because they are open systems. Perhaps he is right but not because of the open system issue but because it is not clear whether the universe can really be modeled as a turing machine. I only wanted to clarify this point and did not claim to have found something new. -Ursprüngliche Nachricht- Von: Vladimir Nesov [mailto:[EMAIL PROTECTED] Gesendet: Samstag, 26. April 2008 20:55 An: agi@v2.listbox.com Betreff: Re: [agi] How general can be and should be AGI? On Sat, Apr 26, 2008 at 9:39 PM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: I think, your argumentation is that an AGI system (e.g. human being) can solve any halting problem because it can change over time by making more and more experiences. But the even the experience making human being can be regarded as a turing machine with a fixed and finite algorithm. It's pretty bald to expect that many people could've been missing such a flaw in your straw man for years, isn't it? -- Vladimir Nesov [EMAIL PROTECTED] --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
On Sat, Apr 26, 2008 at 11:42 PM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: Don't understand your point fully. Perhaps my English is too bad. I have had the impression, that pei wang thought that gödels theorem and the halting problem do not apply for human beings because they are open systems. Perhaps he is right but not because of the open system issue but because it is not clear whether the universe can really be modeled as a turing machine. I only wanted to clarify this point and did not claim to have found something new. My complaint was merely about what I heard as assumption that many people here believe that AGI must be able to learn to start solving halting problems. Your below assertion does have many problems (think of the universe as a finite state machine, or as a fixed Turing machine running forward, or in other way limited in information): On Sat, Apr 26, 2008 at 6:03 PM, Dr. Matthias Heger [EMAIL PROTECTED] wrote: So, with Goedel, the whole universe cannot generate a set of statements about itself which are both complete and without any contradictions. And therefore a machine which is part of the universe cannot have this ability too. -- Vladimir Nesov [EMAIL PROTECTED] --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
2008/4/26 Dr. Matthias Heger [EMAIL PROTECTED]: How general should be AGI? My answer, as *potentially* general as possible. In a similar fashion that a UTM is as potentially as general as possible, but with more purpose. There are plenty of problems you can define that don't need the halting problem to be impossible to solve, e.g. remember a number with more digits that the potential states of the universe. Some other comments. Have you looked at the literature on neuro plasticity? This wired article is a good introduction. http://www.wired.com/wired/archive/15.04/esp_pr.html Although there are more academic papers out there, a google can find them. Will Pearson --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
BillK: MT: So what you must tell me is how your or any geometrical system of analysis is going to be able to take a rorschach and come up similarly with a recognizable object or creature. Bear in mind, your system will be given no initial clues as to what objects or creatures are suitable as potential comparisons. It can by all means have a large set of visual images in memory, as we do. But you must tell me how your system will connect the rorschach with any of those images, such as a bat, - by *geometrical* means. snip This is called Content-based image retrieval (CBIR), also known as query by image content (QBIC) and content-based visual information retrieval (CBVIR) is the application of computer vision to the image retrieval problem, that is, the problem of searching for digital images in large databases. http://en.wikipedia.org/wiki/CBIR This is a hot area of computer research, with many test systems. (see article). Nothing to do with AGI, of course. BillK, CBIR isn't the same as drawing analogies and forming metaphors - or comparing rorschachs to bats. You're saying that analogy and metaphor are not important to AGI? I disagree. I think they're central. CBIR is about retrieving the *same/ v.similar* kinds of objects or shapes (rather than radically different ones that neverthless have some not-obvious similarity, e.g. moustaches and walruses). And no, the ability to analyse images in terms of more or less the same visual elements, like say, the colour red, or the same texture or shape, will NOT solve the problem of analogy/metaphor or comparing rorschachs and bats. May I suggest, BTW, that you really *look* at the problem - put the two images side by side? http://www.desordre.net/textes/bibliotheque/rorschach.jpg http://members.tripod.com/~susano/images/bat2.gif And CBIR isn't really working, is it? (Not, for a second, that it's without its uses). It can't find pictures of dogs can it? You could call that a problem of conceptualisation. Which, if you go into it enough, is deeply related to the problem of comparing rorschachs and bats - because 'dogs' or 'labradors' or pretty well any species come in very diverse, not-so-similar shapes. Thanks for the ref. P.S. Re my negativity, I'm saying broadly that AI and AGI lack certain crucial faculties - you think they have all the faculties they need to succeed? --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] How general can be and should be AGI?
You've missed the point. What a human does in looking at a rorschach is to see - i.e. compare it with - a recognizable object or creature - a bat, for instance, or an ant, or a gargoyle. I didn't miss the point. The standard visual operators are doing exactly the same thing. So what you must tell me is how your or any geometrical system of analysis is going to be able to take a rorschach and come up similarly with a recognizable object or creature. Bear in mind, your system will be given no initial clues as to what objects or creatures are suitable as potential comparisons. It can by all means have a large set of visual images in memory, as we do. But you must tell me how your system will connect the rorschach with any of those images, such as a bat, - by *geometrical* means. Mike, do you know what vector graphics are? Do you understand how comparing vector graphics can lead to exactly such an identification? Why are you asking this question as if this something new or unique? Of course, a geometrical system can be used to *analyse* an individual rorschach into some set of geometric forms - but only by hand, individually, on a one-off basis - and imperfectly. There is no geometrical *formula* for analysing rorschachs, because they can take an unlimited and non-formulaic variety of shapes, just as there is no geometrical formula for analysing the diverse shapes of living creatures, like bats, ants and human beings. So there is, by extension, no geometrical, or indeed any other formulaic means to *compare* a rorschach with any object or creature. There are all sorts of ad hoc algorithms that can replace what you say must be done by hand. The second half of your paragraph is just blatantly incorrect. Nor is there any geometrical means to compare *any* irregular objects - a slug and, say, a human being walking along, a purse, say, and a vagina, a rock and a chair, a moustache and a walrus. Wrong. If you think there is, then you obviously have solved some of the most important, unsolved problems of AGI, such as analogy, metaphor and creativity. How so? Show me *exactly* how they correlate and how if I have solved the one, the other is trivial. You've also turned geometers into designers and artists How so? Since when does decomposition equal good composition? Personally, I am able to analyze art and say what is good and bad. I am not, however, a particularly good artist. Your argument is just plain wrong. AGAIN. So I await your geometric solution to this problem - (a mere statement of principle will do) - with great interest. Well, actually no. Your answer is broadly predictable - you 1) won't have any idea here 2) will have nothing to say to the point and 3) be, as usual, all bark and no bite - all insults and no ideas. Nice ad hominem. Asshole. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
WARNING -- LET'S KEEP THE LIST CIVIL PLEASE ... was Re: [agi] How general can be and should be AGI?
Ummm... just a little note of warning from the list owner. Tintner wrote: So I await your geometric solution to this problem - (a mere statement of principle will do) - with great interest. Well, actually no. Your answer is broadly predictable - you 1) won't have any idea here 2) will have nothing to say to the point and 3) be, as usual, all bark and no bite - all insults and no ideas. Waser wrote: Nice ad hominem. Asshole. Uh, no. Mark, you've been a really valuable contributor to this list for a long period of time. But, this sort of name-calling is just not apropos on this list. Don't do it anymore. Thanks Ben --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
RE: [agi] How general can be and should be AGI?
I assume you are referring to Mike Tintner. As I described a while ago, I *plonk*ed him myself a long time ago, most mail programs have the ability to do that. and it's a good idea to figure out how to do it with your own email program. He does have the ability to point at other thinkers and their papers, such as Lakoff and Barsalou, who have extremely interesting things to say... but his own contributions (beyond citing) to any converation are infuriating., I think it's about time to give up on Mike until he learns to behave again. And you shouldn't use sarcasm -- he just doesn't get it. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
