Re: [agi] AGI Alife
Interesting article: http://www.newscientist.com/article/mg20727723.700-artificial-life-forms-evolve-basic-intelligence.html?page=1 On Sun, Aug 1, 2010 at 3:13 PM, Jan Klauck jkla...@uni-osnabrueck.dewrote: Ian Parker wrote I would like your opinion on *proofs* which involve an unproven hypothesis, I've no elaborated opinion on that. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
This is much more interesting in the context of Evolution than it is for the creation of AGI. Point is that all the things that have ben done would have been done (much more simply in fact) from straightforward narrow programs. However it demonstrates the early multicelluar organisms of the Pre Cambrian and early Cambrian. What AGI is interested in is how *language* evolves. That is to say the last 6 million years or so. We also need a process for creating AGI which is rather more efficient than Evolution. We can't wait that time for something to happen. - Ian Parker On 6 August 2010 19:23, rob levy r.p.l...@gmail.com wrote: Interesting article: http://www.newscientist.com/article/mg20727723.700-artificial-life-forms-evolve-basic-intelligence.html?page=1 On Sun, Aug 1, 2010 at 3:13 PM, Jan Klauck jkla...@uni-osnabrueck.dewrote: Ian Parker wrote I would like your opinion on *proofs* which involve an unproven hypothesis, I've no elaborated opinion on that. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com *agi* | Archives https://www.listbox.com/member/archive/303/=now https://www.listbox.com/member/archive/rss/303/ | Modifyhttps://www.listbox.com/member/?;Your Subscription http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
This is on the surface interesting. But I'm kinda dubious about it. I'd like to know exactly what's going on - who or what (what kind of organism) is solving what kind of problem about what? The exact nature of the problem and the solution, not just a general blurb description. If you follow the link from Kurzweil, you get a really confusing picture/screen. And I wonder whether the real action/problemsolving isn't largely taking place in the viewer/programmer's mind. From: rob levy Sent: Friday, August 06, 2010 7:23 PM To: agi Subject: Re: [agi] AGI Alife Interesting article: http://www.newscientist.com/article/mg20727723.700-artificial-life-forms-evolve-basic-intelligence.html?page=1 On Sun, Aug 1, 2010 at 3:13 PM, Jan Klauck jkla...@uni-osnabrueck.de wrote: Ian Parker wrote I would like your opinion on *proofs* which involve an unproven hypothesis, I've no elaborated opinion on that. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com agi | Archives | Modify Your Subscription --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
Ian Parker wrote I would like your opinion on *proofs* which involve an unproven hypothesis, I've no elaborated opinion on that. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
Adding is simple proving is hard. This is a truism. I would like your opinion on *proofs* which involve an unproven hypothesis, such as Riemann. Hardy and Littlewood proved Goldbach with this assumption. Unfortunately the does not apply. The truth of Goldbach does not imply the Riemann hypothesis. Riemann would be proved if a converse was valid and the theorem proved another way. I am not really arguing deep philosophy, what I am saying is that a non inscrutable system must go to its basic axioms. - Ian Parker On 31 July 2010 00:25, Jan Klauck jkla...@uni-osnabrueck.de wrote: Ian Parker wrote Then define your political objectives. No holes, no ambiguity, no forgotten cases. Or does the AGI ask for our feedback during mission? If yes, down to what detail? With Matt's ideas it does exactly that. How does it know when to ask? You give it rules, but those rules can be somehow imperfect. How are its actions monitored and sanctioned? And hopefully it's clear that we are now far from mathematical proof. No we simply add to the axiom pool. Adding is simple, proving is not. Especially when the rules, goals, and constraints are not arithmetic but ontological and normative statements. Wether by NL or formal system, it's error-prone to specify our knowledge of the world (much of it is implicit) and teach it to the AGI. It's similar to law which is similar to math with referenced axioms and definitions and a substitution process. You often find flaws--most are harmless, some are not. Proofs give us islands of certainty in an explored sea within the ocean of the possible. We end up with heuristics. That's what this discussion is about, when I remember right. :) cu Jan --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
Ian Parker wrote Then define your political objectives. No holes, no ambiguity, no forgotten cases. Or does the AGI ask for our feedback during mission? If yes, down to what detail? With Matt's ideas it does exactly that. How does it know when to ask? You give it rules, but those rules can be somehow imperfect. How are its actions monitored and sanctioned? And hopefully it's clear that we are now far from mathematical proof. No we simply add to the axiom pool. Adding is simple, proving is not. Especially when the rules, goals, and constraints are not arithmetic but ontological and normative statements. Wether by NL or formal system, it's error-prone to specify our knowledge of the world (much of it is implicit) and teach it to the AGI. It's similar to law which is similar to math with referenced axioms and definitions and a substitution process. You often find flaws--most are harmless, some are not. Proofs give us islands of certainty in an explored sea within the ocean of the possible. We end up with heuristics. That's what this discussion is about, when I remember right. :) cu Jan --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
On 28 July 2010 23:09, Jan Klauck jkla...@uni-osnabrueck.de wrote: Ian Parker wrote If we program a machine for winning a war, we must think well what we mean by winning. I wasn't thinking about winning a war, I was much more thinking about sexual morality and men kissing. If we program a machine for doing X, we must think well what we mean by X. Now clearer? Winning a war is achieving your political objectives in the war. Simple definition. Then define your political objectives. No holes, no ambiguity, no forgotten cases. Or does the AGI ask for our feedback during mission? If yes, down to what detail? With Matt's ideas it does exactly that. The axioms which we cannot prove should be listed. You can't prove them. Let's list them and all the assumptions. And then what? Cripple the AGI by applying just those theorems we can prove? That excludes of course all those we're uncertain about. And it's not so much a single theorem that's problematic but a system of axioms and inference rules that changes its properties when you modify it or that is incomplete from the beginning. No we simply add to the axiom pool. *All* I am saying is that we must always have a lemma train taking us to the most fundamental Suppose I say W=AσT4 Now I ask the system to prove this. At the bottom of the lemma trail will be Clifford algebra. This relates Bose Einstein statistics to the spin, in this case of the photon. It is Quantum Mechanics at a very fundamental level. A Fermion has a half in its spin. I can introduce as many axioms as I want. I can say that i = √-1. I can call this statement an axiom, as a counter example of your natural numbers. In constructing Clifford Algebra I make a number of statements. This thinking in terms of axioms I repeat does not limit the power of AGI. If we have a database you could almost say that a lemma trail was in essence trivial. What is does do is invalidate the biological model. *An absolute requirement for AGI is openness.* In other words we must be able to examine the arguments and their validity. Example (very plain just to make it clearer what I'm talking about): The natural numbers N are closed against addition. But N is not closed against subtraction, since n - m 0 where m n. You can prove the theorem that subtracting a positive number from another number decreases it: http://us2.metamath.org:88/mpegif/ltsubpos.html but you can still have a formal system that runs into problems. In the case of N it's missing closedness, i.e., undefined area. Now transfer this simple example to formal systems in general. You have to prove every formal system as it is, not just a single theorem. The behavior of an AGI isn't a single theorem but a system. The heuristics could be tested in an off line system. Exactly. But by definition heuristics are incomplete, their solution space is smaller than the set of all solutions. No guarantee for the optimal solution, just probabilities 1, elaborated hints. Unselfishness going wrong is in fact a frightening thought. It would in AGI be a symptom of incompatible axioms. Which can happen in a complex system. Only if the definitions are vague. I bet against this. Better to have a system based on *democracy* in some form or other. The rules you mention are goals and constraints. But they are heuristics you check during runtime. That is true. Also see above. System cannot be inscruitable. - Ian Parker --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
On 27 July 2010 21:06, Jan Klauck jkla...@uni-osnabrueck.de wrote: Second observation about societal punishment eliminating free loaders. The fact of the matter is that *freeloading* is less of a problem in advanced societies than misplaced unselfishness. Fact of the matter, hm? Freeloading is an inherent problem in many social configurations. 9/11 brought down two towers, freeloading can bring down an entire country. There are very considerable knock on costs. There is the mushrooming cost of security This manifests itself in many ways. There is the cost of disruption to air travel. If someone rides on a plane without a ticket no one's life is put at risk. There are the military costs, it costs $1m per year to keep a soldier in Afghanistan. I don't know how much a Taliban fighter costs, but it must be a lot less. Clearly any reduction in these costs would be welcomed. If someone were to come along in the guise of social simulation and offer a reduction in these costs the research would pay for itself many times over. What *you* are interested in. This may be a somewhat unpopular thing to say, but money *is* important. Matt Mahoney has costed his view of AGI. I say that costs must be recoverable as we go along. Matt, don't frighten people with a high estimate of cost. Frighten people instead with the bill they are paying now for dumb systems. simulations seem :- 1) To be better done by Calculus. You usually use both, equations and heuristics. It depends on the problem, your resources, your questions, the people working with it a.s.o. That is the way things should be done. I agree absolutely. We could in fact take steepest descent (Calculus) and GAs and combine them together in a single composite program. This would in fact be quite a useful exercise. We would also eliminate genes that simply dealt with Calculus and steepest descent. I don't know whether it is useful to think in topological terms. - Ian Parker --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
One last point. You say freeloading can cause o society to disintegrate. One society that has come pretty damn close to disintegration is Iraq. The deaths in Iraq were very much due to sectarian blood letting. Unselfishness if you like. Would that the Iraqis (and Afghans) were more selfish. - Ian Parker --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
Ian Parker wrote: Matt Mahoney has costed his view of AGI. I say that costs must be recoverable as we go along. Matt, don't frighten people with a high estimate of cost. Frighten people instead with the bill they are paying now for dumb systems. It is not my intent to scare people out of building AGI, but rather to be realistic about its costs. Building machines that do what we want is a much harder problem than building intelligent machines. Machines surpassed human intelligence 50 years ago. But getting them to do useful work is still a $60 trillion per year problem. It's going to happen, but not as quickly as one might hope. -- Matt Mahoney, matmaho...@yahoo.com From: Ian Parker ianpark...@gmail.com To: agi agi@v2.listbox.com Sent: Wed, July 28, 2010 6:54:05 AM Subject: Re: [agi] AGI Alife On 27 July 2010 21:06, Jan Klauck jkla...@uni-osnabrueck.de wrote: Second observation about societal punishment eliminating free loaders. The fact of the matter is that *freeloading* is less of a problem in advanced societies than misplaced unselfishness. Fact of the matter, hm? Freeloading is an inherent problem in many social configurations. 9/11 brought down two towers, freeloading can bring down an entire country. There are very considerable knock on costs. There is the mushrooming cost of security This manifests itself in many ways. There is the cost of disruption to air travel. If someone rides on a plane without a ticket no one's life is put at risk. There are the military costs, it costs $1m per year to keep a soldier in Afghanistan. I don't know how much a Taliban fighter costs, but it must be a lot less. Clearly any reduction in these costs would be welcomed. If someone were to come along in the guise of social simulation and offer a reduction in these costs the research would pay for itself many times over. What you are interested in. This may be a somewhat unpopular thing to say, but money is important. Matt Mahoney has costed his view of AGI. I say that costs must be recoverable as we go along. Matt, don't frighten people with a high estimate of cost. Frighten people instead with the bill they are paying now for dumb systems. simulations seem :- 1) To be better done by Calculus. You usually use both, equations and heuristics. It depends on the problem, your resources, your questions, the people working with it a.s.o. That is the way things should be done. I agree absolutely. We could in fact take steepest descent (Calculus) and GAs and combine them together in a single composite program. This would in fact be quite a useful exercise. We would also eliminate genes that simply dealt with Calculus and steepest descent. I don't know whether it is useful to think in topological terms. - Ian Parker --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com agi | Archives | Modify Your Subscription --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
Ian Parker wrote There are the military costs, Do you realize that you often narrow a discussion down to military issues of the Iraq/Afghanistan theater? Freeloading in social simulation isn't about guys using a plane for free. When you analyse or design a system you look for holes in the system that allow people to exploit it. In complex systems that happens often. Most freeloading isn't much of a problem, just friction, but some have the power to damage the system too much. You have that in the health system, social welfare, subsidies and funding, the usual moral hazard issues in administration, services a.s.o. To come back to AGI: when you hope to design, say, a network of heterogenous neurons (taking Linas' example) you should be interested in excluding mechanisms that allow certain neurons to consume resources without delivering something in return because of the way resource allocation is organized. These freeloading neurons could go undetected for a while but when you scale the network up or confront it with novel inputs they could make it run slow or even break it. If someone were to come along in the guise of social simulation and offer a reduction in these costs the research would pay for itself many times over. SocSim research into peace and conflict studies isn't new. And some people in the community work on the Iraq/Afghanistan issue (for the US). That is the way things should be done. I agree absolutely. We could in fact take steepest descent (Calculus) and GAs and combine them together in a single composite program. This would in fact be quite a useful exercise. Just a note: Social simulation is not so much about GAs. You use agent systems and equation systems. Often you mix both in that you define the agent's behavior and the environment via equations, let the sim run and then describe the results in statistical terms or with curve fitting in equations again. One last point. You say freeloading can cause o society to disintegrate. One society that has come pretty damn close to disintegration is Iraq. The deaths in Iraq were very much due to sectarian blood letting. Unselfishness if you like. Unselfishness gone wrong is a symptom, not a cause. The causes for failed states are different. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
Unselfishness gone wrong is a symptom. I think that this and all the other examples should be cautionary for anyone who follows the biological model. Do we want a system that thinks the way we do. Hell no! What we would want in a *friendly* system would be a set of utilitarian axioms. That would immediately make it think differently from us. We certainly would not want a system which would arrest men kissing on a park bench. In other words we would not want a system which was axiomatically righteous. It is also important that AGI is fully axiomatic and proves that 1+1=2 by set theory, as Russell did. This immediately takes it out of the biological sphere. We will need morality to be axiomatically defined. Unselfishness going wrong is in fact a frightening thought. It would in AGI be a symptom of incompatible axioms. In humans it is a real problem and it should tell us that AGI cannot and should not be biologically based. On 28 July 2010 15:59, Jan Klauck jkla...@uni-osnabrueck.de wrote: Ian Parker wrote There are the military costs, Do you realize that you often narrow a discussion down to military issues of the Iraq/Afghanistan theater? Freeloading in social simulation isn't about guys using a plane for free. When you analyse or design a system you look for holes in the system that allow people to exploit it. In complex systems that happens often. Most freeloading isn't much of a problem, just friction, but some have the power to damage the system too much. You have that in the health system, social welfare, subsidies and funding, the usual moral hazard issues in administration, services a.s.o. To come back to AGI: when you hope to design, say, a network of heterogenous neurons (taking Linas' example) you should be interested in excluding mechanisms that allow certain neurons to consume resources without delivering something in return because of the way resource allocation is organized. These freeloading neurons could go undetected for a while but when you scale the network up or confront it with novel inputs they could make it run slow or even break it. In point of fact we can look at this another way. Lets dig a little bit deeperhttp://sites.google.com/site/aitranslationproject/computergobbledegook. If we have one AGI system we can have 2 (or 3 even, automatic landing in fog is a triplex system). Suppose system A is monitoring system B. If system Bs resources are being used up A can shut down processes in A. I talked about computer gobledegook. I also have the feeling that with AGI we should be able to get intelligible advice (in NL) about what was going wrong. For this reason it would not be possible to overload AGI. I have the feeling that perhaps one aim in AGI should be user friendly systems. One product is in fact a form filler. As far as society i concerned I think this all depends on how resource limited we are. In a resource limited society freeloading is the biggest issue. In our society violence in all its forms is the big issue. One need not go to Iraq or Afghanistan for examples. There are plenty in ordinary crime. Happy slapping, domestic violence, violence against children. If the people who wrote computer viruses stole a large sum of money, what they did would, to me at any rate, be more forgiveable. People take a delight in wrecking things for other people, while not stealing very much themselves. Iraq, Afghanistan and suicide murder is really simply an extreme example of this. Why I come back to it is that the people feel they are doing Allah's will. Happy slappers usually say they have nothing better to do. The fundamental fact about Western crime is that very little of it is to do with personal gain or greed. If someone were to come along in the guise of social simulation and offer a reduction in these costs the research would pay for itself many times over. SocSim research into peace and conflict studies isn't new. And some people in the community work on the Iraq/Afghanistan issue (for the US). That is the way things should be done. I agree absolutely. We could in fact take steepest descent (Calculus) and GAs and combine them together in a single composite program. This would in fact be quite a useful exercise. Just a note: Social simulation is not so much about GAs. You use agent systems and equation systems. Often you mix both in that you define the agent's behavior and the environment via equations, let the sim run and then describe the results in statistical terms or with curve fitting in equations again. One last point. You say freeloading can cause o society to disintegrate. One society that has come pretty damn close to disintegration is Iraq. The deaths in Iraq were very much due to sectarian blood letting. Unselfishness if you like. Unselfishness gone wrong is a symptom, not a cause. The causes for failed states are different. Axiomatic contradiction. Cannot occur in a mathematical system. -
Re: [agi] AGI Alife
Ian Parker wrote What we would want in a *friendly* system would be a set of utilitarian axioms. If we program a machine for winning a war, we must think well what we mean by winning. (Norbert Wiener, Cybernetics, 1948) It is also important that AGI is fully axiomatic and proves that 1+1=2 by set theory, as Russell did. Quoting the two important statements from http://en.wikipedia.org/wiki/Principia_Mathematica#Consistency_and_criticisms Gödel's first incompleteness theorem showed that Principia could not be both consistent and complete. and Gödel's second incompleteness theorem shows that no formal system extending basic arithmetic can be used to prove its own consistency. So in effect your AGI is either crippled but safe or powerful but potentially behaves different from your axiomatic intentions. We will need morality to be axiomatically defined. As constraints, possibly. But we can only check the AGI in runtime for certain behaviors (i.e., while it's active), but we can't prove in advance whether it will break the constraints or not. Get me right: We can do a lot with such formal specifications and we should do them where necessary or appropriate, but we have to understand that our set of guaranteed behavior is a proper subset of the set of all possible behaviors the AGI can execute. It's heuristics in the end. Unselfishness going wrong is in fact a frightening thought. It would in AGI be a symptom of incompatible axioms. Which can happen in a complex system. Suppose system A is monitoring system B. If system Bs resources are being used up A can shut down processes in A. I talked about computer gobledegook. I also have the feeling that with AGI we should be able to get intelligible advice (in NL) about what was going wrong. For this reason it would not be possible to overload AGI. This isn't going to guarantee that system A, B, etc. behave in all ways as intended, except they are all special purpose systems (here: narrow AI). If A, B etc. are AGIs, then this checking is just an heuristic, no guarantee or proof. In a resource limited society freeloading is the biggest issue. All societies are and will be constrained by limited resources. The fundamental fact about Western crime is that very little of it is to do with personal gain or greed. Not that sure whether this statement is correct. It feels wrong from what I know about human behavior. Unselfishness gone wrong is a symptom, not a cause. The causes for failed states are different. Axiomatic contradiction. Cannot occur in a mathematical system. See above... --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
On 28 July 2010 19:56, Jan Klauck jkla...@uni-osnabrueck.de wrote: Ian Parker wrote What we would want in a *friendly* system would be a set of utilitarian axioms. If we program a machine for winning a war, we must think well what we mean by winning. I wasn't thinking about winning a war, I was much more thinking about sexual morality and men kissing. Winning a war is achieving your political objectives in the war. Simple definition. (Norbert Wiener, Cybernetics, 1948) It is also important that AGI is fully axiomatic and proves that 1+1=2 by set theory, as Russell did. Quoting the two important statements from http://en.wikipedia.org/wiki/Principia_Mathematica#Consistency_and_criticisms Gödel's first incompleteness theorem showed that Principia could not be both consistent and complete. and Gödel's second incompleteness theorem shows that no formal system extending basic arithmetic can be used to prove its own consistency. So in effect your AGI is either crippled but safe or powerful but potentially behaves different from your axiomatic intentions. You have to state what your axioms are. Gödel's theorem does indeed state that. You do have to make therefore some statements which are unprovable. What I was in fact thinking in terms of was something like Mizar. Mathematics starts off with simple ideas. The axioms which we cannot prove should be listed. You can't prove them. Let's list them and all the assumptions. If we have a Mizar proof we assume things, and argue the case for a theorem on what we have assumed. What you should be able to do is get from the ideas of Russell and Bourbaki to something really meaty like Fermat's Last Theorem, or the Riemann hypothesis. The organization of Mizar (and Alcor which is a front end) is very much a part of AGI. Alcor has in fact to do a similar job to Google in terms of a search for theorems. Mizar though is different from Google in that we have lemmas. You prove something by linking the lemmas up. Suppose I were to search for Riemann Hypothesis. Alcor should give me all the theorems that depend on it. It should tell me about the density of primes. It should tell me about the Goldbach conjecture, proved by Hardy and Littlewood to depend on Riemann. Google is a step towards AGI. An Alcor which could produce chains of argument and find lemmas would be a big step to AGI. Could Mizar contain knowledge which was non mathematical? In a sense it already can. Mizar will contain Riemanian differential geometry. This is simply a piece of pure maths. I am allowed to make a conjecture, an axiom if you like that Riemann's differential geometry is in the shape of General relativity the way in which the Universe works. I have stated this as an unproven assertion, one that has been constantly verified experimentally but unproven in the mathematical universe. We will need morality to be axiomatically defined. As constraints, possibly. But we can only check the AGI in runtime for certain behaviors (i.e., while it's active), but we can't prove in advance whether it will break the constraints or not. Get me right: We can do a lot with such formal specifications and we should do them where necessary or appropriate, but we have to understand that our set of guaranteed behavior is a proper subset of the set of all possible behaviors the AGI can execute. It's heuristics in the end. The heuristics could be tested in an off line system. Unselfishness going wrong is in fact a frightening thought. It would in AGI be a symptom of incompatible axioms. Which can happen in a complex system. Only if the definitions are vague. The definition of happiness is vague. Better to have a system based on *democracy* in some form or other. The beauty of Matt's system is that we would remain ultimately in charge of the system. We make rules such as no imprisonment without trial, minimum of laws restriction personal freedom (men kissing), separation of powers in the judiciary and executive and the reolution of disputed without violence. These are I repeat *not* fundamental philosophical principles but rules which our civilization has devised and have been found to work. I have mentioned before that we could have more than 1 AGI system. All the *derived* principles would be tested off line on another AGI system. Suppose system A is monitoring system B. If system Bs resources are being used up A can shut down processes in A. I talked about computer gobledegook. I also have the feeling that with AGI we should be able to get intelligible advice (in NL) about what was going wrong. For this reason it would not be possible to overload AGI. This isn't going to guarantee that system A, B, etc. behave in all ways as intended, except they are all special purpose systems (here: narrow AI). If A, B etc. are AGIs, then this checking is just an heuristic, no guarantee or proof. In a resource limited society freeloading is the biggest
Re: [agi] AGI Alife
Ian Parker wrote If we program a machine for winning a war, we must think well what we mean by winning. I wasn't thinking about winning a war, I was much more thinking about sexual morality and men kissing. If we program a machine for doing X, we must think well what we mean by X. Now clearer? Winning a war is achieving your political objectives in the war. Simple definition. Then define your political objectives. No holes, no ambiguity, no forgotten cases. Or does the AGI ask for our feedback during mission? If yes, down to what detail? The axioms which we cannot prove should be listed. You can't prove them. Let's list them and all the assumptions. And then what? Cripple the AGI by applying just those theorems we can prove? That excludes of course all those we're uncertain about. And it's not so much a single theorem that's problematic but a system of axioms and inference rules that changes its properties when you modify it or that is incomplete from the beginning. Example (very plain just to make it clearer what I'm talking about): The natural numbers N are closed against addition. But N is not closed against subtraction, since n - m 0 where m n. You can prove the theorem that subtracting a positive number from another number decreases it: http://us2.metamath.org:88/mpegif/ltsubpos.html but you can still have a formal system that runs into problems. In the case of N it's missing closedness, i.e., undefined area. Now transfer this simple example to formal systems in general. You have to prove every formal system as it is, not just a single theorem. The behavior of an AGI isn't a single theorem but a system. The heuristics could be tested in an off line system. Exactly. But by definition heuristics are incomplete, their solution space is smaller than the set of all solutions. No guarantee for the optimal solution, just probabilities 1, elaborated hints. Unselfishness going wrong is in fact a frightening thought. It would in AGI be a symptom of incompatible axioms. Which can happen in a complex system. Only if the definitions are vague. I bet against this. Better to have a system based on *democracy* in some form or other. The rules you mention are goals and constraints. But they are heuristics you check during runtime. --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
I spent a while back in the 90s trying to make AGI and alife converge, before establishing to my satisfaction the approach is a dead end: we will never have anywhere near enough computing power to make alife evolve significant intelligence (the only known success took 4 billion years on a planetary sized nanocomputer network, after all), even if we could set up just the right selection pressures, which we can't. On Tue, Jul 27, 2010 at 4:23 AM, Linas Vepstas linasveps...@gmail.com wrote: I saw the following post from Antonio Alberti, on the linked-in discussion group: ALife and AGI Dear group participants. The relation among AGI and ALife greatly interests me. However, too few recent works try to relate them. For exemple, many papers presented in AGI-09 (http://agi-conf.org/2009/) are about program learning algorithms (combining evolutionary learning and analytical learning). In AGI 2010, virtual pets have been presented by Ben Goertzel and are also another topic of this forum. There are other approaches in AGI that uses some digital evolutionary approach for AGI. For me it is a clear clue that both are related in some instance. By ALife I mean the life-as-it-could-be approach (not simulate, but to use digital environment to evolve digital organisms using digital evolution (faster than Natural one - see http://www.hplusmagazine.com/articles/science/stephen-hawking-%E2%80%9Chumans-have-entered-new-stage-evolution%E2%80%9D). So, I would like to propose some discussion topics regarding ALIfe and AGI: 1) What is the role of Digital Evolution (and ALife) in the AGI context? 2) Is it possible that some aspects of AGI could self-emerge from the digital evolution of intelligent autonomous agents? 3) Is there any research group trying to converge both approaches? Best Regards, and my reply was below: For your question 3), I have no idea. For question 1) I can't say I've ever heard of anyone talk about this. For question 2), I imagine the answer is yes, although the boundaries between what's Alife and what's program learning (for example) may be blurry. So, imagine, for example, a population of many different species of neurons (or should I call them automata? or maybe I should call them virtual ants?) Most of the individuals have only a few friends (a narrow social circle) -- the friendship relationship can be viewed as an axon-dendrite connection -- these friendships are semi-stable; they evolve over time, and the type quality of information exchanged in a friendship also varies. Is a social network of friends able to solve complex problems? The answer is seemingly yes, if the individuals are digital models of neurons. (To carry analogy further: different species of individuals would be analogous to different types of neurons e.g. purkinje cells vs pyramid cells vs granular vs. motor neurons. Individuals from one species may tend to be very gregarious, while those from other species might be generally xenophobic. etc.) I have no clue if anyone has ever explored genetic algorithms or related alife algos, factored together with the individuals being involved in a social network (with actual information exchange between friends). No clue as to how natural/artificial selection should work. Do anti-social individuals have a possibly redeeming role w.r.t. the organism as a whole? Do selection pressures on individuals (weak individuals are cullled) destroy social networks? Do such networks automatically evolve altruism, because a working social network with weak, altruistically-supported individuals is better than a shredded, dysfunctional social network consisting of only strong individuals? Dunno. Seems like there could be many many interesting questions. I'd be curious about the answers to Antonio's questions ... --linas --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
Linas Vepstas wrote First my answers to Antonio: 1) What is the role of Digital Evolution (and ALife) in the AGI context? The nearest I can come up with is Goertzel's virtual pre-school idea, where the environment is given and the proto-AGI learns within it. It's certainly possible to place such a proto-AGI into an evolving environment. I'm not sure how helpful this is, since now we also need to make sense of the evolving environment in order to assess what the agent does. But that's far from the synthetic life approach, where environment and agents are usually not that much pre-defined. And from those synth. approaches I know about, they're mostly concerned with replicating natural evolution, adaption, self-organization a.s.o. Some look into the emergence and evolution of cooperation, but that's often very low level and more interested in general properties; far from AGI. 2) Is it possible that some aspects of AGI could self-emerge from the digital evolution of intelligent autonomous agents? I guess it's possible. But I guess one won't come up with a mechanism that works in an AGI system but with interesting properties of an AGI system. Most intelligent agents are faked, not really cognitive or so. In a simulation you see how agents develop/select strategies and what works in an (evolutionary) environment. Like (wild idea now) the ability to assign parts of its cognitive capacity to memory or processing depending on the environmental context (more memory in unchanging and more processing in changing environments). Those properties could be integrated later as a detail of a bigger framework. 3) Is there any research group trying to converge both approaches? My best ad-hoc idea is to scan through the last year's alife conference program, look for papers that are promising, contact the authors and ask whether they are into AGI or know people who are. http://www.ecal2009.org/documents/ECAL2009_program.pdf One of the topics was artificial consciousness and I saw several papers going into this direction, often indirectly. Like the Swarm Cognition and Artificial Life paper on p.34 or the first poster on p.47. Now to Linas' part: Seems like there could be many many interesting questions. Many of these are specialized issues that are researched in alife but more in social simulation. The Journal of Artificial Societies and Social Simulation http://jasss.soc.surrey.ac.uk/JASSS.html is a good starting point if anyone is interested. cu Jan --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
I did take a look at the journal. There is one question I have with regard to the assumptions. Mathematically the number of prisoners in Prisoner's dilemma cooperating or not reflects the prevalence of cooperators or non cooperators present. Evolution *should* tend to Von Neumann's zero sum condition. This is an example of Calculus solving a problem far neater and more elegantly than GAs which should only be used where there is no good or obvious Calculus solution. This is my first observation. Second observation about societal punishment eliminating free loaders. The fact of the matter is that *freeloading* is less of a problem in advanced societies than misplaced unselfishness. The 9/11 hijackers performed the most unselfish and unfreeloading acts. Hope I am not accused of glorifying terrorism! How fundamental an issue is this? It is fundamental in that simulations seem :- 1) To be better done by Calculus. 2) Not to be useful in providing simulations of things we are interested in. Neither of these two is necessarily the case. We could in fact simulate opinion formation by social interaction. There there would be no clear cut Calculus outcome. The third observation is that Google is itself a GA. It uses popular appeal in its page ranking systems. This is relevant to Matt's ideas. You can, for example, string programs or other entities together. Of course to do this association one needs Natural Language. You will also need NL in stetting up and describing any process of opinion formation. This is the great unsolved problem. In fact any system not based on NL, but based on a analogue response is Calculus describable. - Ian Parker On 27 July 2010 14:00, Jan Klauck jkla...@uni-osnabrueck.de wrote: Seems like there could be many many interesting questions. Many of these are specialized issues that are researched in alife but more in social simulation. The Journal of Artificial Societies and Social Simulation http://jasss.soc.surrey.ac.uk/JASSS.html is a good starting point if anyone is interested. cu Jan --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
I think I should say that for a problem to be suitable for GAs the space in which it is embedded has to be non linear. Otherwise we have an easy Calculus solution. http://www.springerlink.com/content/h46r77k291rn/?p=bfaf36a87f704d5cbcb66429f9c8a808pi=0 is described a fair number of such systems. - Ian Parker --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Alife
Evolving AGI via an Alife approach would be possible, but would likely take many orders of magnitude more resources than engineering AGI... I worked on Alife years ago and became frustrated that the artificial biology and artificial chemistry one uses is never as fecund as the real thing We don't understand which aspects of bio and chem are really important for the evolution of complex structures. So, approaching AGI via Alife just replaces one complex set of confusions with another ;-) ... I think that releasing some well-engineered AGI systems in an Alife type environment, and letting them advance and evolve further, would be an awesome experiment, though ;) -- Ben G On Mon, Jul 26, 2010 at 11:23 PM, Linas Vepstas linasveps...@gmail.com wrote: I saw the following post from Antonio Alberti, on the linked-in discussion group: ALife and AGI Dear group participants. The relation among AGI and ALife greatly interests me. However, too few recent works try to relate them. For exemple, many papers presented in AGI-09 (http://agi-conf.org/2009/) are about program learning algorithms (combining evolutionary learning and analytical learning). In AGI 2010, virtual pets have been presented by Ben Goertzel and are also another topic of this forum. There are other approaches in AGI that uses some digital evolutionary approach for AGI. For me it is a clear clue that both are related in some instance. By ALife I mean the life-as-it-could-be approach (not simulate, but to use digital environment to evolve digital organisms using digital evolution (faster than Natural one - see http://www.hplusmagazine.com/articles/science/stephen-hawking-%E2%80%9Chumans-have-entered-new-stage-evolution%E2%80%9D). So, I would like to propose some discussion topics regarding ALIfe and AGI: 1) What is the role of Digital Evolution (and ALife) in the AGI context? 2) Is it possible that some aspects of AGI could self-emerge from the digital evolution of intelligent autonomous agents? 3) Is there any research group trying to converge both approaches? Best Regards, and my reply was below: For your question 3), I have no idea. For question 1) I can't say I've ever heard of anyone talk about this. For question 2), I imagine the answer is yes, although the boundaries between what's Alife and what's program learning (for example) may be blurry. So, imagine, for example, a population of many different species of neurons (or should I call them automata? or maybe I should call them virtual ants?) Most of the individuals have only a few friends (a narrow social circle) -- the friendship relationship can be viewed as an axon-dendrite connection -- these friendships are semi-stable; they evolve over time, and the type quality of information exchanged in a friendship also varies. Is a social network of friends able to solve complex problems? The answer is seemingly yes, if the individuals are digital models of neurons. (To carry analogy further: different species of individuals would be analogous to different types of neurons e.g. purkinje cells vs pyramid cells vs granular vs. motor neurons. Individuals from one species may tend to be very gregarious, while those from other species might be generally xenophobic. etc.) I have no clue if anyone has ever explored genetic algorithms or related alife algos, factored together with the individuals being involved in a social network (with actual information exchange between friends). No clue as to how natural/artificial selection should work. Do anti-social individuals have a possibly redeeming role w.r.t. the organism as a whole? Do selection pressures on individuals (weak individuals are cullled) destroy social networks? Do such networks automatically evolve altruism, because a working social network with weak, altruistically-supported individuals is better than a shredded, dysfunctional social network consisting of only strong individuals? Dunno. Seems like there could be many many interesting questions. I'd be curious about the answers to Antonio's questions ... --linas --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC CTO, Genescient Corp Vice Chairman, Humanity+ Advisor, Singularity University and Singularity Institute External Research Professor, Xiamen University, China b...@goertzel.org I admit that two times two makes four is an excellent thing, but if we are to give everything its due, two times two makes five is sometimes a very charming thing too. -- Fyodor Dostoevsky --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your
[agi] AGI Alife
I saw the following post from Antonio Alberti, on the linked-in discussion group: ALife and AGI Dear group participants. The relation among AGI and ALife greatly interests me. However, too few recent works try to relate them. For exemple, many papers presented in AGI-09 (http://agi-conf.org/2009/) are about program learning algorithms (combining evolutionary learning and analytical learning). In AGI 2010, virtual pets have been presented by Ben Goertzel and are also another topic of this forum. There are other approaches in AGI that uses some digital evolutionary approach for AGI. For me it is a clear clue that both are related in some instance. By ALife I mean the life-as-it-could-be approach (not simulate, but to use digital environment to evolve digital organisms using digital evolution (faster than Natural one - see http://www.hplusmagazine.com/articles/science/stephen-hawking-%E2%80%9Chumans-have-entered-new-stage-evolution%E2%80%9D). So, I would like to propose some discussion topics regarding ALIfe and AGI: 1) What is the role of Digital Evolution (and ALife) in the AGI context? 2) Is it possible that some aspects of AGI could self-emerge from the digital evolution of intelligent autonomous agents? 3) Is there any research group trying to converge both approaches? Best Regards, and my reply was below: For your question 3), I have no idea. For question 1) I can't say I've ever heard of anyone talk about this. For question 2), I imagine the answer is yes, although the boundaries between what's Alife and what's program learning (for example) may be blurry. So, imagine, for example, a population of many different species of neurons (or should I call them automata? or maybe I should call them virtual ants?) Most of the individuals have only a few friends (a narrow social circle) -- the friendship relationship can be viewed as an axon-dendrite connection -- these friendships are semi-stable; they evolve over time, and the type quality of information exchanged in a friendship also varies. Is a social network of friends able to solve complex problems? The answer is seemingly yes, if the individuals are digital models of neurons. (To carry analogy further: different species of individuals would be analogous to different types of neurons e.g. purkinje cells vs pyramid cells vs granular vs. motor neurons. Individuals from one species may tend to be very gregarious, while those from other species might be generally xenophobic. etc.) I have no clue if anyone has ever explored genetic algorithms or related alife algos, factored together with the individuals being involved in a social network (with actual information exchange between friends). No clue as to how natural/artificial selection should work. Do anti-social individuals have a possibly redeeming role w.r.t. the organism as a whole? Do selection pressures on individuals (weak individuals are cullled) destroy social networks? Do such networks automatically evolve altruism, because a working social network with weak, altruistically-supported individuals is better than a shredded, dysfunctional social network consisting of only strong individuals? Dunno. Seems like there could be many many interesting questions. I'd be curious about the answers to Antonio's questions ... --linas --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com