Re: [agi] Approximations of Knowledge
On 6/23/08, J. Andrew Rogers <[EMAIL PROTECTED]> wrote: > > Or it could simply mean that the vast majority of programmers and software > monkeys are mediocre at best such that the handful of people you will meet > with deep talent won't constitute a useful sample size. Hell, even Brooks > suggested as much and he was charitable. In all my years in software, I've > only met a small number of people who were unambiguously wicked smart when > it came to software, and while none of them could be confused with a > completely mundane person they also did not have many other traits in common > (though I will acknowledge they tend to rational and self-analytical to a > degree that is rare in most people though this is not a trait exclusive to > these people). Of course, *my* sample size is also small and so it does not > count for much. I completely agree with all of the above, though it says nothing relevant to the point that I was trying to make. That point was that we and presumably our AGIs will use our experience to focus inquiry in complex situations. That these focused efforts fail more often than they succeed is good, compared with the disastrous alternative of failing 99.99% of the time because our inquiries are NOT focused. Again, as you apparently missed it on my previous email - what would you suggest as an alternative? > Similarly, over the course of >100 projects... >> > > Eh? Over 100 projects? These were either very small projects or you are > older than Methuselah. Both are correct. Also, I had many fewer employers, as I had a LOT of repeat business. These would sometimes bring me in for a couple of weeks of "shock treatment" when they felt it was needed. Steve Richfield --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
On Jun 23, 2008, at 7:53 PM, Steve Richfield wrote: Andy, The use of diminutives is considered rude in many parts of anglo- culture if the individual does not use it to identify themselves, though I realize it is common practice in some regions of the US. When in doubt, use the given form. This is a PERFECT post, because it so perfectly illustrates a particular point of detachment from reality that is common among AGIers. In the real world we do certain things to achieve a good result, but when we design politically correct AGIs, we banish the very logic that allows us to function. For example, if you see a black man walking behind you at night, you rightly worry, but if you include that in your AGI design, you would be dismissed as a racist. You have clearly confused me with someone else. Effectively solving VERY VERY difficult problems, like why a particular corporation is failing after other experts have failed, is a multiple-step process that starts with narrowing down the vast field of possibilities. As others have already pointed out here, this is often done in a rather summary and non-probabilistic way. Perhaps all of the really successful programmers that you have known have had long hair, so if the programming is failing and the programmer has short hair, then maybe there is an attitude issue to look into. Of course this does NOT necessarily mean that there is any linkage at all - just another of many points to focus some attention to. Or it could simply mean that the vast majority of programmers and software monkeys are mediocre at best such that the handful of people you will meet with deep talent won't constitute a useful sample size. Hell, even Brooks suggested as much and he was charitable. In all my years in software, I've only met a small number of people who were unambiguously wicked smart when it came to software, and while none of them could be confused with a completely mundane person they also did not have many other traits in common (though I will acknowledge they tend to rational and self-analytical to a degree that is rare in most people though this is not a trait exclusive to these people). Of course, *my* sample size is also small and so it does not count for much. Similarly, over the course of >100 projects... Eh? Over 100 projects? These were either very small projects or you are older than Methuselah. I've worked on a lot of projects, but nowhere near 100 and I was a consultant for many years. J. Andrew Rogers --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
Andy, This is a PERFECT post, because it so perfectly illustrates a particular point of detachment from reality that is common among AGIers. In the real world we do certain things to achieve a good result, but when we design politically correct AGIs, we banish the very logic that allows us to function. For example, if you see a black man walking behind you at night, you rightly worry, but if you include that in your AGI design, you would be dismissed as a racist. Effectively solving VERY VERY difficult problems, like why a particular corporation is failing after other experts have failed, is a multiple-step process that starts with narrowing down the vast field of possibilities. As others have already pointed out here, this is often done in a rather summary and non-probabilistic way. Perhaps all of the really successful programmers that you have known have had long hair, so if the programming is failing and the programmer has short hair, then maybe there is an attitude issue to look into. Of course this does NOT necessarily mean that there is any linkage at all - just another of many points to focus some attention to. Similarly, over the course of >100 projects I have developed a long list of "rules" that help me find the problems with a tractable amount of effort. No, I don't usually tell others my poorly-formed rules because they prove absolutely NOTHING, only focus further effort. I have a special assortment of rules to apply whenever God is mentioned. After all, not everyone thinks that God has the same motivations, so SOME approach is needed to "paradigm shift" one person's statements to be able to be understood by another person. The posting you responded to was expressing one such rule. That having been said... On 6/22/08, J. Andrew Rogers <[EMAIL PROTECTED]> wrote: > > > Somewhere in the world, there is a PhD chemist and a born-again Christian > on another mailing list "...the project had hit a serious snag, and so the > investors brought in a consultant that would explain why the project was > broken by defectively reasoning about dubious generalizations he pulled out > of his ass..." Of course I don't make any such (I freely admit to dubious) generalizations to investors. However, I immediately drill down to find out exactly why THEY SAY that they didn't stop and reconsider their direction when it should have been obvious that things had gone off track. When I hear about how God just couldn't have led them astray, I quote what they said in my report and suggest that perhaps the problem is that God isn't also underwriting the investment with limitless funds. How would YOU (or your AGI) handle such situations? Would you (or your AGI) ignore past empirical evidence because of lack of proof or political incorrectness? Steve Richfield --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
Jim Bromer wrote: Loosemore said, "It is very important to understand that the paper I wrote was about the methodology of AGI research, not about specific theories/models/systems within AGI. It is about the way that we come up with ideas for systems and the way that we explore those systems, not about the content of anyone's particular ideas." And Abram said, "A revised version of my argument would run something like this. As the approximation problem gets more demanding, it gets more difficult to devise logical heuristics. Increasingly, we must rely on intuitions tested by experiments. There then comes a point when making the distinction between the heuristic and the underlying search becomes unimportant; the method is all heuristic, so to speak. At this point we are simply using "messy" methods," I wondered if Abram was talking about the way an AI program should work or the way research into AI should work, or the way AI programs and research into AI should work? Jim Bromer I interpreted him (see parallel post) to be referring still to the question of how to deal with planning systems, where there is a formalism (the logic substructure) which cannot be allowed to run its methods to completion (because they would take too long) and which therefore has to use "approximation methods", or heuristics, to guess which are the most likely best planning choices. When the system is required to do more real-world-type performance (as in an AGI, rather than a narrow AI) it's behavior will be dominated by the heuristics. He then went on to talk about methodology: do we just use intuitions to pick heuristics, or do we make the methodology more systematic by engaging in automatic searches of the space of possible heuristics? My perspective on that question would back up one step: if it is a complex system we are dealing with, we should have been using systematic, automatic searches of the design space BEFORE, when we were choosing whether or not to do planning with a Logic+Heuristics design! But of course, that would be wildly, extravagantly infeasible. So, instead, I propose to start from a basic design that is as similar as possible to the human design, and then do our systematic, automatic search (of the space of mechanism-designs) in an outward direction from that human-cognition baseline. If intelligence involves even a small amount of complexity, it could well be that this is the only feasible way to ever get an intelligence up and running. Treat it, in other words, as a calculus of variations problem. Richard Loosemore. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
Abram Demski wrote: Thanks for the comments. My replies: It does happen to be the case that I believe that logic-based methods are mistaken, but I could be wrong about that, and it could turn out that the best way to build an AGI is with a completely logic-based AGI, along with just one small mechanism that was Complex. Logical methods are quite Complex. This was part of my point. Logical deduction in any sufficiently complicated formalism satisfies both types of global-local disconnect that I mentioned (undecidability, and computational irreducibility). If this were not the case, it seems your argument would be much less appealing. (In particular, there would be one less argument for the mind being complex; we could not say "logic has some subset of the mind's capabilities; a brute-force theorem prover is a complex system; therefore the mind is probably a complex system.") Okay, I made a mistake in my choice of words (I knew it when I wrote them, but neglected to go back and correct!). I did not mean to imply that I *require* some complexity in an AGI formalism, and that finding some complexity would be a good thing, end of story, problem solved, etc. So for example, you are correct to point out that most 'logical' systems do exhibit complexity, provided they do something realistically approximating intelligence. Instead, what I meant to say was that we are not setting up our research procedures to cope with the complexity. So, it might turn out that a good, robust AGI can be built with something like a regular logic-based formalism, BUT with just a few small aspects that are complex but unfortunately we are currently not able to discover what those complex parts should be like, because our current methodology is to use blind hunch and intuition (i.e. heuristics that "look" as though the will work). Going back to your planning system example, it might be the case that only one choice of heuristic control mechanism will actually make a given logical formalism converge on fully intelligent behavior, but there might be 10^100 choices of possible control mechanism, and our current method for searching through the possibilities is to use intuition to pick likely candidates. The point here is that a small amount of the factors that give rise to complexity can actualy have a massive effect on the behavior of the system, but people are today acting as if a small amount of complexity-inducing characteristics means a small amount of unpredictability in the behavior. This is simply not the case. Similarly, you suggest that I "have an image of an AGI that is built out of totally dumb pieces, with intelligence emerging unexpectedly." Some people have suggested that that is my view of AGI, but whether or not those people are correct in saying that [aside: they are not!] Apologies. But your arguments do appear to point in that direction. In your original blog post, also, you mention the way that AGI planning The problem is that you have portrayed the distinction between 'pure' logical mechanisms and 'messy' systems that have heuristics riding on their backs, as equivalent to a distinction that you thought I was making between non-complex and complex AGI systems. I hope you can see now that this is not what I was trying to argue. You are right, this characterization is quite bad. I think that is part of what was making me uneasy about my conclusion. My intention was not that approximation should always equal a logical search with messy heuristics stacked upon it. In fact, I had two conflicting images in mind:use -A logical search with logical heuristics (such as greedy methods for NP-complete problems, which are guaranteed to be fairly near optimal) -A "messy" method (such as a neural net or swarm) that somehow gives you an answer without precise logic A revised version of my argument would run something like this. As the approximation problem gets more demanding, it gets more difficult to devise logical heuristics. Increasingly, we must rely on intuitions tested by experiments. There then comes a point when making the distinction between the heuristic and the underlying search becomes unimportant; the method is all heuristic, so to speak. At this point we are simply using "messy" methods. Ah, I agree completely here. We are taling about a Wag The Dog scenario, where everyone focusses on the pristine beauty of the logical formalism, but turns a blind eye to the (assumed-to-be) trivial heuristic control mechanisms but in the end it is the heuristic control mechanism that is responsible for almost all of the actual behavior. I'm still not really satisfied, though, because I would personally stop at the stage when the heuristic started to get messy, and say, "The problem is starting to become AI-complete, so at this point I should include a meta-level search to find a good heuristic for me, rather than trying to hard-code one..." And at t
Re: [agi] Equivalent ..P.S.I just realised - how can you really understand what I'm talking about - without supplementary images/evidence?
I just realised - how can you really understand what I'm talking about - without supplementary images/evidence? So here's simple evidence - look at the following foto - and note that you can distinguish each individual in it immediately. And you can only do it imagistically. No maths, no language, no algebraic variables, no programming languages can tell you what makes each one of those people individual/ different. Just images. So uniquely powerful. http://www.cdomusic.com/MilitaryImages/AALL%20CASUALTIES%20IN%20ONE%20POSTER.jpg Every thing in the world is individual. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 11:57 PM, Mike Tintner <[EMAIL PROTECTED]> wrote: > Oh yes, it can be proven. It requires an extended argument to do so > properly, which I won't attempt here. Fair enough, I'd be interested to see your attempted proof if you ever get it written up. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
Russell:quite a few very smart people (myself among them) have tried hard to design something that could enhance its intelligence divorced from the real world, and all such attempts have failed. Obviously I can't _prove_ the impossibility of this - in the same way that I can't prove the impossibility of summoning demons by chanting the right phrases in Latin; you can always say, well maybe there's some incantation nobody has yet tried. Oh yes, it can be proven. It requires an extended argument to do so properly, which I won't attempt here. But it all comes down, if you think about it, to different forms of sign/representation. The AGI-ers who think knowledge can be superaccelerated are almost exclusively talking about knowledge in the form of symbols - logical, mathematical, linguistic. When you or I talk about gathering evidence and personal experience, we are talking about knowledge gathered in the form of sensory images - and I am also talking about "embodied" images - which involve your whole body (that's what mirror neurons are referring to - when you mimic someone or something, you do it with your whole body, not just your senses). The proof lies in the direction of thinking of the world as consisting of "bodies" - and then asking: what can and can't the different kinds of sign: symbols - words/numbers/ algebraic-logical variables, - and then image schemas - geometric figures etc. and then images - sensory/ photographs/movies/ etc - tell you and show you of bodies? Each form of sign/representation has strictly v. limited powers , and can only show certain dimensions of bodies. All the symbols and schemas in existence cannot tell you what Russell Wallace or Vladimir Nesov look like - i.e. cannot show you their distinctive, individual bodies. Only images (or, if you like, "evidence") can do that - and do it in a second. (And that can be proven, scientifically). And since the real world consists, in the final analysis, of nothing but individual bodies like Russell and Vlad, each of which are different from each other - even that ipod over there is actually different from this ipod here, - then you'd better have images if you want to be intelligent about the real world of real individuals, and be able to deal with all their idiosyncrasies - or make fresh generalisations about them. Which is why evolution went to the extraordinary trouble of founding real AGI's on the continuous set of moving images we call consciousness - in order to be able to deal with the real world of individuals, and not just the rational world of abstract general classes, we call logic, maths and language.* But, as I said, this requires an extended argument to demonstrate properly. But, yes, it can be proven. *In case that's confusing, language and logic can refer to individuals like "Russell Wallace" - but only in general terms. They can't show what distinguishes those individuals. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Tue, Jun 24, 2008 at 1:29 AM, Russell Wallace <[EMAIL PROTECTED]> wrote: > On Mon, Jun 23, 2008 at 8:48 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: >> There are only evolution-built animals, which is a very limited >> repertoir of intelligences. You are saying that if no apple tastes >> like a banana, therefore no fruit tastes like a banana, even banana. > > I'm saying if no fruit anyone has ever tasted confers magical powers, > and theory says fruit can't do so, and there's no evidence whatsoever > that it can, then we should accept that eating fruit does not confer > magical powers. Yes, we are discussing the theory that banana taste (magical power) doesn't exist. But this theory mustn't consist in merely asserting that there are no precedents, and pass the absence of precedents for evidence. If there is more to the theory, what is the idea in hand-picking this weak point? >> Whether a design is possible or not, you expect to see the same >> result, if it was never attempted. And so, the absence of an >> implementation of design that was never attempted is not evidence of >> impossibility of design. > > But it has been attempted. I cited not only biological evolution and > learning within the lifetime of individuals, but all fields of science > and engineering - including AI, where quite a few very smart people > (myself among them) have tried hard to design something that could > enhance its intelligence divorced from the real world, and all such > attempts have failed. I have only a very vague idea about what you mean by "intelligence divorced from the real world". Without justification, it looks like a scapegoat. > Obviously I can't _prove_ the impossibility of this - in the same way > that I can't prove the impossibility of summoning demons by chanting > the right phrases in Latin; you can always say, well maybe there's > some incantation nobody has yet tried. Maybe there is, but we don't have any hints about the processes that would produce such an effect, much less a prototype demon-summoning device at any level of obfuscation, so there is little prior in that endeavor. Whereas with intelligence, we have a prototype and plenty of theory that seems to grope for the process, but not quite capture it. > But here's a question for you: Is the possibility of intelligence > enhancement in a vacuum a matter of absolute faith, or is there some > point at which you would accept it's impossible after all? If the > latter, when will you accept its futility? Ten years from now? Twenty? > Thirty? As I said earlier, I don't see any inherent dichotomies between the search for fundamental process and understanding of existing biological brains. It doesn't need to be a political decision, if at some point the brain-inspired technology turns out to be a better path, or more likely, informs the theory, let's take it. For now, it looks like cliff-jumping. -- Vladimir Nesov [EMAIL PROTECTED] http://causalityrelay.wordpress.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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
Loosemore said, "It is very important to understand that the paper I wrote was about the methodology of AGI research, not about specific theories/models/systems within AGI. It is about the way that we come up with ideas for systems and the way that we explore those systems, not about the content of anyone's particular ideas." And Abram said, "A revised version of my argument would run something like this. As the approximation problem gets more demanding, it gets more difficult to devise logical heuristics. Increasingly, we must rely on intuitions tested by experiments. There then comes a point when making the distinction between the heuristic and the underlying search becomes unimportant; the method is all heuristic, so to speak. At this point we are simply using "messy" methods," I wondered if Abram was talking about the way an AI program should work or the way research into AI should work, or the way AI programs and research into AI should work? Jim Bromer - Original Message From: Abram Demski <[EMAIL PROTECTED]> To: agi@v2.listbox.com Sent: Monday, June 23, 2008 3:11:16 PM Subject: Re: [agi] Approximations of Knowledge Thanks for the comments. My replies: > It does happen to be the case that I > believe that logic-based methods are mistaken, but I could be wrong about > that, and it could turn out that the best way to build an AGI is with a > completely logic-based AGI, along with just one small mechanism that was > Complex. Logical methods are quite Complex. This was part of my point. Logical deduction in any sufficiently complicated formalism satisfies both types of global-local disconnect that I mentioned (undecidability, and computational irreducibility). If this were not the case, it seems your argument would be much less appealing. (In particular, there would be one less argument for the mind being complex; we could not say "logic has some subset of the mind's capabilities; a brute-force theorem prover is a complex system; therefore the mind is probably a complex system.") > Similarly, you suggest that I "have an image of an AGI that is built out of > totally dumb pieces, with intelligence emerging unexpectedly." Some people > have suggested that that is my view of AGI, but whether or not those people > are correct in saying that [aside: they are not!] Apologies. But your arguments do appear to point in that direction. > In your original blog post, also, you mention the way that AGI planning > The problem is that you have portrayed the > distinction between 'pure' logical mechanisms and 'messy' systems that have > heuristics riding on their backs, as equivalent to a distinction that you > thought I was making between non-complex and complex AGI systems. I hope > you can see now that this is not what I was trying to argue. You are right, this characterization is quite bad. I think that is part of what was making me uneasy about my conclusion. My intention was not that approximation should always equal a logical search with messy heuristics stacked upon it. In fact, I had two conflicting images in mind:use -A logical search with logical heuristics (such as greedy methods for NP-complete problems, which are guaranteed to be fairly near optimal) -A "messy" method (such as a neural net or swarm) that somehow gives you an answer without precise logic A revised version of my argument would run something like this. As the approximation problem gets more demanding, it gets more difficult to devise logical heuristics. Increasingly, we must rely on intuitions tested by experiments. There then comes a point when making the distinction between the heuristic and the underlying search becomes unimportant; the method is all heuristic, so to speak. At this point we are simply using "messy" methods. I'm still not really satisfied, though, because I would personally stop at the stage when the heuristic started to get messy, and say, "The problem is starting to become AI-complete, so at this point I should include a meta-level search to find a good heuristic for me, rather than trying to hard-code one..." > Finally, I should mention one general misunderstanding about mathematics. > This argument has a superficial similarity to Godel's theorem, but you > should not be deceived by that. Godel was talking about formal deductive > systems, and the fact that there are unreachable truths within such systems. > My argument is about the feasibility of scientific discovery, when applied > to systems of different sorts. These are two very different domains. I think it is fair to say that I accounted for this. In particular, I said: "It's this second kind of irreducibility, computational irreducibility, that I see as more relevant to AI." (Actually, I do see Godel's theorem as relevant to AI; I should have been more specific and said "relevant to AI's global-local disconnect".) --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS
Re: [agi] Coin-flipping duplicates (was: Breaking Solomonoff induction (really))
--- On Mon, 6/23/08, Kaj Sotala <[EMAIL PROTECTED]> wrote: > a) Perform the experiment several times. If, on any of the trials, > copies are created, then have all of them partake in the next trial as > well, flipping a new coin and possibly being duplicated again (and > quickly leading to an exponentially increasing number of copies). > Carry out enough trials to eliminate the effect of random chance. > Since every agent is flipping a fair coin each time, by the time you > finish running the trials, all of them will remember seeing a roughly > equal amount of heads and tails. Knowing this, a rational agent should > anticipate this result, and not a 99% ratio. That is my meaning. But you can run a simulation yourself. The agents that see heads get copied, so you have more agents remembering heads than remembering tails. -- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 8:48 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: > There are only evolution-built animals, which is a very limited > repertoir of intelligences. You are saying that if no apple tastes > like a banana, therefore no fruit tastes like a banana, even banana. I'm saying if no fruit anyone has ever tasted confers magical powers, and theory says fruit can't do so, and there's no evidence whatsoever that it can, then we should accept that eating fruit does not confer magical powers. > Whether a design is possible or not, you expect to see the same > result, if it was never attempted. And so, the absence of an > implementation of design that was never attempted is not evidence of > impossibility of design. But it has been attempted. I cited not only biological evolution and learning within the lifetime of individuals, but all fields of science and engineering - including AI, where quite a few very smart people (myself among them) have tried hard to design something that could enhance its intelligence divorced from the real world, and all such attempts have failed. Obviously I can't _prove_ the impossibility of this - in the same way that I can't prove the impossibility of summoning demons by chanting the right phrases in Latin; you can always say, well maybe there's some incantation nobody has yet tried. But here's a question for you: Is the possibility of intelligence enhancement in a vacuum a matter of absolute faith, or is there some point at which you would accept it's impossible after all? If the latter, when will you accept its futility? Ten years from now? Twenty? Thirty? --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
Vlad, You seem to be arguing in a logical vacuum in denying the essential nature of evidence to most real-world problem-solving. Let's keep it real, bro. Science - bear in mind science deals with every part of the world - from the cosmos to the earth to living organisms, animals, humans, societies etc. Which branch of science can solve problems about the world without evidence and physically interacting with the subject matter? Technology - which branch of technology can solve problems without evidence & interacting with machines and artefacts and the real world? Ditto: which branch of AI or AGI can solve problems without interacting with real-world :computers? (Some purely logical, mathematical problems yes, but overwhelmingly, no). Real-world technology - i.e. business etc - which branch can solve problems without interacting with real products and real customers? History/journalism ...etc. etc. If you think AGI's can somehow magically transcend the requirement to have physical, personal experience and evidence of a subject in order to solve problems about that subject, you must explain how. Preferably with reference to the real world, and not just by using logical argument. As Zeno's paradox shows, logic can prove anything, no matter how absurd. Science and real world intelligence, which are tied to evidence, can't. Evidence is an indication that depends on the referred event: evidence is there when referred event is there, but evidence is not there when refereed event is absent. And if the referred thing (entities acquiring intelligence from static corpus in the absence of environment) existed we would expect to see it happening, if (as I claim) it does not exist then we would expect to see all intelligence-acquiring entities needing interaction with an environment; we observe the latter, which by the above criterion is evidence for my theory. There are only evolution-built animals, which is a very limited repertoir of intelligences. You are saying that if no apple tastes like a banana, therefore no fruit tastes like a banana, even banana. Whether a design is possible or not, you expect to see the same result, if it was never attempted. And so, the absence of an implementation of design that was never attempted is not evidence of impossibility of design. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 9:35 PM, Russell Wallace <[EMAIL PROTECTED]> wrote: > On Mon, Jun 23, 2008 at 5:58 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: > >> Evidence is an indication that depends on the >> referred event: evidence is there when referred event is there, but >> evidence is not there when refereed event is absent. > > And if the referred thing (entities acquiring intelligence from static > corpus in the absence of environment) existed we would expect to see > it happening, if (as I claim) it does not exist then we would expect > to see all intelligence-acquiring entities needing interaction with an > environment; we observe the latter, which by the above criterion is > evidence for my theory. > There are only evolution-built animals, which is a very limited repertoir of intelligences. You are saying that if no apple tastes like a banana, therefore no fruit tastes like a banana, even banana. Whether a design is possible or not, you expect to see the same result, if it was never attempted. And so, the absence of an implementation of design that was never attempted is not evidence of impossibility of design. -- Vladimir Nesov [EMAIL PROTECTED] http://causalityrelay.wordpress.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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
Thanks for the comments. My replies: > It does happen to be the case that I > believe that logic-based methods are mistaken, but I could be wrong about > that, and it could turn out that the best way to build an AGI is with a > completely logic-based AGI, along with just one small mechanism that was > Complex. Logical methods are quite Complex. This was part of my point. Logical deduction in any sufficiently complicated formalism satisfies both types of global-local disconnect that I mentioned (undecidability, and computational irreducibility). If this were not the case, it seems your argument would be much less appealing. (In particular, there would be one less argument for the mind being complex; we could not say "logic has some subset of the mind's capabilities; a brute-force theorem prover is a complex system; therefore the mind is probably a complex system.") > Similarly, you suggest that I "have an image of an AGI that is built out of > totally dumb pieces, with intelligence emerging unexpectedly." Some people > have suggested that that is my view of AGI, but whether or not those people > are correct in saying that [aside: they are not!] Apologies. But your arguments do appear to point in that direction. > In your original blog post, also, you mention the way that AGI planning > The problem is that you have portrayed the > distinction between 'pure' logical mechanisms and 'messy' systems that have > heuristics riding on their backs, as equivalent to a distinction that you > thought I was making between non-complex and complex AGI systems. I hope > you can see now that this is not what I was trying to argue. You are right, this characterization is quite bad. I think that is part of what was making me uneasy about my conclusion. My intention was not that approximation should always equal a logical search with messy heuristics stacked upon it. In fact, I had two conflicting images in mind:use -A logical search with logical heuristics (such as greedy methods for NP-complete problems, which are guaranteed to be fairly near optimal) -A "messy" method (such as a neural net or swarm) that somehow gives you an answer without precise logic A revised version of my argument would run something like this. As the approximation problem gets more demanding, it gets more difficult to devise logical heuristics. Increasingly, we must rely on intuitions tested by experiments. There then comes a point when making the distinction between the heuristic and the underlying search becomes unimportant; the method is all heuristic, so to speak. At this point we are simply using "messy" methods. I'm still not really satisfied, though, because I would personally stop at the stage when the heuristic started to get messy, and say, "The problem is starting to become AI-complete, so at this point I should include a meta-level search to find a good heuristic for me, rather than trying to hard-code one..." > Finally, I should mention one general misunderstanding about mathematics. > This argument has a superficial similarity to Godel's theorem, but you > should not be deceived by that. Godel was talking about formal deductive > systems, and the fact that there are unreachable truths within such systems. > My argument is about the feasibility of scientific discovery, when applied > to systems of different sorts. These are two very different domains. I think it is fair to say that I accounted for this. In particular, I said: "It's this second kind of irreducibility, computational irreducibility, that I see as more relevant to AI." (Actually, I do see Godel's theorem as relevant to AI; I should have been more specific and said "relevant to AI's global-local disconnect".) --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 5:58 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: > On Mon, Jun 23, 2008 at 8:32 PM, Russell Wallace >> Why do you think that? All the evidence is to the contrary - the >> examples we have of figuring out efficient learning, from evolution to >> childhood play to formal education and training to science to hardward >> and software engineering, do not work with just a static corpus. > > It is not evidence. Yes it is. > Evidence is an indication that depends on the > referred event: evidence is there when referred event is there, but > evidence is not there when refereed event is absent. And if the referred thing (entities acquiring intelligence from static corpus in the absence of environment) existed we would expect to see it happening, if (as I claim) it does not exist then we would expect to see all intelligence-acquiring entities needing interaction with an environment; we observe the latter, which by the above criterion is evidence for my theory. > What would you > expect to see, depending on correctness of your assumption? Literally, > it translates to animals having a phase where they sit cross-legged > and meditate on accumulated evidence, until they gain enlightenment, > become extremely efficient learners and launch Singularity... ...er, I think there's a miscommunication here - I'm claiming this is _not_ possible. I thought you were claiming it is? --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 8:32 PM, Russell Wallace <[EMAIL PROTECTED]> wrote: > On Mon, Jun 23, 2008 at 5:22 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: >> But it can just work with a static corpus. When you need to figure out >> efficient learning, you only need to know a little about the overall >> structure of your data (which can be described by a reasonably small >> number of exemplars), you don't need much of the data itself. > > Why do you think that? All the evidence is to the contrary - the > examples we have of figuring out efficient learning, from evolution to > childhood play to formal education and training to science to hardward > and software engineering, do not work with just a static corpus. > It is not evidence. Evidence is an indication that depends on the referred event: evidence is there when referred event is there, but evidence is not there when refereed event is absent. What would you expect to see, depending on correctness of your assumption? Literally, it translates to animals having a phase where they sit cross-legged and meditate on accumulated evidence, until they gain enlightenment, become extremely efficient learners and launch Singularity... Evolution just didn't figure it out, just like it didn't figure out transistors, and had to work with legacy 100Hz neurons. -- Vladimir Nesov [EMAIL PROTECTED] http://causalityrelay.wordpress.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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 5:22 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: > But it can just work with a static corpus. When you need to figure out > efficient learning, you only need to know a little about the overall > structure of your data (which can be described by a reasonably small > number of exemplars), you don't need much of the data itself. Why do you think that? All the evidence is to the contrary - the examples we have of figuring out efficient learning, from evolution to childhood play to formal education and training to science to hardward and software engineering, do not work with just a static corpus. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 7:52 PM, Russell Wallace <[EMAIL PROTECTED]> wrote: > On Mon, Jun 23, 2008 at 4:34 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: >> On Mon, Jun 23, 2008 at 6:52 PM, Russell Wallace >>> Indeed, but becoming more efficient at processing evidence is >>> something that requires being embedded in the environment to which the >>> evidence pertains. >> >> Why is that? > > For the reason I explained earlier. Suppose program A generates > candidate programs B1, B2... that are conjectured to be more efficient > at processing evidence. It can't just compare their processing of > evidence with the correct version, because if it knew the correct > results in all cases, it would already be that efficient itself. It > has to try them out. > But it can just work with a static corpus. When you need to figure out efficient learning, you only need to know a little about the overall structure of your data (which can be described by a reasonably small number of exemplars), you don't need much of the data itself. -- Vladimir Nesov [EMAIL PROTECTED] http://causalityrelay.wordpress.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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
William Pearson wrote: While SIAI fills that niche somewhat, it concentrates on the Intelligence explosion scenario. Is there a sufficient group of researchers/thinkers with a shared vision of the future of AI coherent enough to form an organisation? This organisation would discus, explore and disseminate what can be done to make the introduction as painless as possible. The base beliefs shared between the group would be something like - The entities will not have goals/motivations inherent to their form. That is robots aren't likely to band together to fight humans, or try to take over the world for their own means. These would have to be programmed into them, as evolution has programmed group loyalty and selfishness into humans. - The entities will not be capable of fully wrap around recursive self-improvement. They will improve in fits and starts in a wider economy/ecology like most developments in the world * - The goals and motivations of the entities that we will likely see in the real world will be shaped over the long term by the forces in the world, e.g. evolutionary, economic and physics. Basically an organisation trying to prepare for a world where AIs aren't sufficiently advanced technology or magic genies, but still dangerous and a potentially destabilising world change. Could a coherent message be articulated by the subset of the people that agree with these points. Or are we all still too fractured? Will Pearson * I will attempt to give an inside view of why I take this view, at a later date. The Bulletin of the Atomic Scientists is an organization that started with a precise idea, based on extremely well-established theory, of the dangers of nuclear technology. At this time there is nothing like a coherent theory from which we could draw conclusions about the (possible) dangers of AGI. Such an organization would be pointless. It is bad enough that SIAI is 50% community mouthpiece and 50% megaphone for Yudkowsky's ravings. More mouthpieces we don't need. Richard Loosemore --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 4:34 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: > On Mon, Jun 23, 2008 at 6:52 PM, Russell Wallace >> Indeed, but becoming more efficient at processing evidence is >> something that requires being embedded in the environment to which the >> evidence pertains. > > Why is that? For the reason I explained earlier. Suppose program A generates candidate programs B1, B2... that are conjectured to be more efficient at processing evidence. It can't just compare their processing of evidence with the correct version, because if it knew the correct results in all cases, it would already be that efficient itself. It has to try them out. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
Since combinatorial search problems are so common to artificial intelligence, it has obvious applications. If such an algorithm can be made, it seems like it could be used *everywhere* inside an AGI: deduction (solve for cases consistent with constraints), induction (search for the best model), planning... Particularly if there is a generalization to soft constraint problems. On 6/22/08, Jim Bromer <[EMAIL PROTECTED]> wrote: > Abram, > I did not group you with "probability buffs". One of the errors I feel that > writers make when their field is controversial is that they begin > representing their own opinions from the vantage of countering critics. > Unfortunately, I am one of those writers, (or perhaps I am just projecting). > But my comment about the probability buffs wasn't directed toward you, I > was just using it as an exemplar (of something or another). > > Your comments seem to make sense to me although I don't know where you are > heading. You said: > "what should be hoped for is convergence to (nearly) correct models of > (small parts of) the universe. So I suppose that rather than asking for > "meaning" in a fuzzy logic, I should be asking for clear accounts of > convergence properties..." > > When you have to find a way to tie together components of knowledge together > you typically have to achieve another kind of convergence. Even if these > 'components' of knowledge are reliable, they cannot usually be converged > easily due to the complexity that their interrelations with other kinds of > knowledge (other 'components' of knowledge) will cause. > > To follow up on what I previously said, if my logic program works it will > mean that I can combine and test logical formulas of up to a few hundred > distinct variables and find satisfiable values for these combinations in a > relatively short period of time. I think this will be an important method > to test whether AI can be advanced by advancements in handling complexity > even though some people do not feel that logical methods are appropriate to > use on multiple source complexity. As you seem to appreciate, logic can > still be brought to to the field even though it is not a purely logical game > that is to be played. > > When I begin to develop some simple theories about a subject matter, I will > typically create hundreds of minor variations concerning those theories over > a period of time. I cannot hold all those variations of the conjecture in > consciousness at any one moment, but I do feel that they can come to mind in > response to a set of conditions for which that particular set of variations > was created for. So while a simple logical theory (about some subject) may > be expressible with only a few terms, when you examine all of the possible > variations that can be brought into conscious consideration in response to a > particular set of stimuli, I think you may find that the theories could be > more accurately expressed using hundreds of distinct logical values. > > If this conjecture of mine turns out to be true, and if I can actually get > my new logical methods to work, then I believe that this new range of > logical methods may show whether advancements in complexity can make a > difference to AI even if its application does not immediately result in > human level of intelligence. > > Jim Bromer > > > - Original Message > From: Abram Demski <[EMAIL PROTECTED]> > To: agi@v2.listbox.com > Sent: Sunday, June 22, 2008 4:38:02 PM > Subject: Re: [agi] Approximations of Knowledge > > Well, since you found my blog, you probably are grouping me somewhat > with the "probability buffs". I have stated that I will not be > interested in any other fuzzy logic unless it is accompanied by a > careful account of the meaning of the numbers. > > You have stated that it is unrealistic to expect a logical model to > reflect the world perfectly. The intuition behind this seems clear. > Instead, what should be hoped for is convergence to (nearly) correct > models of (small parts of) the universe. So I suppose that rather than > asking for "meaning" in a fuzzy logic, I should be asking for clear > accounts of convergence properties... but my intuition says that from > clear meaning, everything else follows. > > > > > > > --- > 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
Abram Demski wrote: To be honest, I am not completely satisfied with my conclusion on the post you refer to. I'm not so sure now that the fundamental split between logical/messy methods should occur at the line between perfect & approximate methods. This is one type of messiness, but one only. I think you are referring to a related but different messiness: not knowing what kind of environment your AI is dealing with. Since we don't know which kinds of models will fit best with the world, we should (1) trust our intuitions to some extent, and (2) try things and see how well they work. This is as Loosemore suggests. On the other hand, I do not want to agree with Loosemore too strongly. Mathematics and mathematical proof is a very important tool, and I feel like he wants to reject it. His image of an AGI seems to be a system built up out of totally dumb pieces, with intelligence emerging unexpectedly. Mine is a system built out of somewhat smart pieces, cooperating to build somewhat smarter pieces, and so on. Each piece has provable smarts. Okay, let me try to make some kind of reply to your comments here and in your original blog post. It is very important to understand that the paper I wrote was about the methodology of AGI research, not about specific theories/models/systems within AGI. It is about the way that we come up with ideas for systems and the way that we explore those systems, not about the content of anyone's particular ideas. So, in the above text you refer to a split between logical and messy methods - now, it may well be that my paper would lead someone to embrace 'messy' methods and reject 'logical' ones, but that is a side effect of the argument, not the argument itself. It does happen to be the case that I believe that logic-based methods are mistaken, but I could be wrong about that, and it could turn out that the best way to build an AGI is with a completely logic-based AGI, along with just one small mechanism that was Complex. That would be perfectly consistent with my argument (though a little surprising, for other reasons). Similarly, you suggest that I "have an image of an AGI that is built out of totally dumb pieces, with intelligence emerging unexpectedly." Some people have suggested that that is my view of AGI, but whether or not those people are correct in saying that [aside: they are not!], that does not relate to the argument I presented, because it is all about specific AGI design preferences, whereas the thing that I have called the "Complex Systems Problem" is fairly neutral on most design decisions. In your original blog post, also, you mention the way that AGI planning mechanisms can be built in such a way that they contain a logical substrate, but with heuristics that force the systems to make 'sub-optimal' choices. This is a specific instance of a more general design pattern: logical engines that have 'inference control mechanisms' riding on their backs, preventing them from deducing everything in the universe whilst trying to come to a simple decision. The problem is that you have portrayed the distinction between 'pure' logical mechanisms and 'messy' systems that have heuristics riding on their backs, as equivalent to a distinction that you thought I was making between non-complex and complex AGI systems. I hope you can see now that this is not what I was trying to argue. My target would be the methodologies that people use to decide such questions as which heuristics to using in a planning mechanism, whether the representation used by the planning mechanism can co-exist with the learning mechanisms, and so on. Now, having said all of that, what does the argument actually say, and does it make *any* claims at all about what sort of content to put in an AGI design? The argument says that IF intelligent systems belong to the 'complex systems' class, THEN a it would be a dreadful mistake to use a certain type of scientific or engineering approach to build intelligent systems. I tried to capture this with an analogy at one point: if you we John Horton Conway, sitting down on Day 1 of his project to find a cellular automaton with certain global properties, you would not be able to use any standard scientific, engineering or mathematical tools to discover the rules that should go into your system - you would, in fact, have no option but to try rules at random until you found rules that gave the global behavior that you desired. My point was that a modified form of that same problem (that inability to use our scientific intuitions to just go from a desired global behavior to the mechanisms that will generate that global behavior) could apply to the question of building an AGI. I do not suggest that the problem will manifest itself in exactly the same way (it is not that we would make zero progress with current techniques, and have to use completely random trial and error, like Conway had to), bu
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 6:52 PM, Russell Wallace <[EMAIL PROTECTED]> wrote: > On Mon, Jun 23, 2008 at 3:43 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: >> We are very inefficient in processing evidence, there is "plenty of >> room at the bottom" in this sense alone. Knowledge doesn't come from >> just feeding the system with data - try to read machine learning >> textbooks to a chimp, nothing will stick. > > Indeed, but becoming more efficient at processing evidence is > something that requires being embedded in the environment to which the > evidence pertains. Why is that? -- Vladimir Nesov [EMAIL PROTECTED] http://causalityrelay.wordpress.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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
2008/6/23 Vladimir Nesov <[EMAIL PROTECTED]>: > On Mon, Jun 23, 2008 at 12:50 AM, William Pearson <[EMAIL PROTECTED]> wrote: >> 2008/6/22 Vladimir Nesov <[EMAIL PROTECTED]>: >> >>> >>> Two questions: >>> 1) Do you know enough to estimate which scenario is more likely? >> >> Well since intelligence explosions haven't happened previously in our >> light cone, it can't be a simple physical pattern, so I think >> non-exploding intelligences have the evidence for being simpler on >> their side. > > This message that I'm currently writing hasn't happened previously in > out light code. By your argument, it is evidence for it being more > difficult to write, than to recreate life on Earth and human > intellect, which is clearly false, for all practical purposes. You > should state that argument more carefully, in order for it to make > sense. If your message was an intelligent entity then you would have a point. I'm looking at classes of technologies and their natural or current human created analogues. Let me give you an example. You have two people claiming to be able to give you an improved TSP solver. One person claims to be able to do all examples in polynomial time the other simply has a better algorithm which can do certain types of graphs in polynomial time, but resorts to exponential time for random graphs. Which would you consider more likely if neither of them have detailed proofs and why? > >> So we might find them more easily. I also think I have >> solid reasoning to think intelligence exploding is unlikely, which >> requires paper length rather than post length. So it I think I do, but >> should I trust my own rationality? > > But not too much, especially when the argument is not technical (which > is clearly the case for questions such as this one). The question is one of theoretical computer science and should be able to be decided as well as the resolution to the halting problem. I'm leaning towards something like Russell Wallace's resolution, but there maybe some complications when you have a program that learns from the environment. I would like to see it done in formally at some point. > If argument is > sound, you should be able to convince seed AI crowd too Since the concept is their idea they have to be the ones to define it. They won't accept any arguments against it otherwise. They haven't as yet formally defined it, or if they have I haven't seen it. > I agree, but it works only if you know that the answer is correct, and > (which you didn't address and which is critical for these issues) you > won't build a doomsday machine as a result of your efforts, even if > this particular path turns out to be more feasible. I don't think a doomsday machine is possible. But considering I would be doing my best to make the system incapable of modifying it's own source code *in the fashion that eliezer wants/is afraid of* anyway, I am not too worried. See http://www.sl4.org/archive/0606/15131.html 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 3:43 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote: > We are very inefficient in processing evidence, there is "plenty of > room at the bottom" in this sense alone. Knowledge doesn't come from > just feeding the system with data - try to read machine learning > textbooks to a chimp, nothing will stick. Indeed, but becoming more efficient at processing evidence is something that requires being embedded in the environment to which the evidence pertains. A chimp did not acquire the ability to read textbooks by sitting in a cave and pondering deep thoughts for a million years. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 5:22 PM, Russell Wallace <[EMAIL PROTECTED]> wrote: > If we step back and think about it, we really knew this already. In > every case where humans, machines or biological systems exhibit > anything that could be called an intelligence improvement - biological > evolution, a child learning to talk, a scientific community improving > its theories, engineers building better aeroplanes, programmers > improving their software - it involves feedback from the environment. > The mistake of trying to reach truth by pure armchair thought was > understandable in ancient Greece. We now know better. > We are very inefficient in processing evidence, there is "plenty of room at the bottom" in this sense alone. Knowledge doesn't come from just feeding the system with data - try to read machine learning textbooks to a chimp, nothing will stick. Intelligence is, among other things, an ability to absorb the data and use it to deftly manipulate the world to your ends, by nudging it here and there. -- Vladimir Nesov [EMAIL PROTECTED] http://causalityrelay.wordpress.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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
Russell:The mistake of trying to reach truth by pure armchair thought was understandable in ancient Greece. We now know better.So attractive as the image of a Transcendent Power popping out of a basement may be to us geeks, it doesn't have anything to do with reality. Making smarter machines in the real world is, like every other engineering activity, a process that has to take place _in_ the real world Just so. I called it the Bookroom Fantasy, (you're almost calling it the Armchair Fallacy), and it does go back philosophically to the Greeks. It all depends on what the Greeks started - the era of rationality, (in the technical sense of the rational sign systems of logic, maths, and, to an extent, language). In rational systems it IS possible to reach truth to a great extent by pure armchair thought - but only truths about rational systems themselves. And you "geeks" (your word) don't seem to have noticed that these systems, while extremely valuable, are only used in strictly limited ways in the real world, & real-world problem solving - and actually pride themselves on being somewhat divorced from reality. "I like mathematics because it is not human and has nothing particular to do with this planet or with the whole accidental universe . Bertrand Russell Mathematics may be defined as the subject in which we never know what we are talking about, nor whether what we are saying is true. Bertrand Russell As far as the laws of mathematics refer to reality, they are not certain; and as far as they are certain, they do not refer to reality. Einstein" The fantasy of super-accelerating intelligence is based on such a simplistic armchair fallacy. And it's ironic because it's cropping up just as the era of rationality is ending. I haven't seen its equivalent, though, in any other area of our culture besides AGI. Roboticists don't seem to have it. What's replacing rationality? I'm still thinking about the best term. I think it's probably *creativity*. The rational era believed in humans as rational animals using pure reason - and especially rational systems - to think about the world. The new creative era is recognizing that thinking about the world, or indeed anything, involves Reason + Emotion + Imagination [Reflective] + Enactment/Embodied Thought + Imagination[Direct Sensory] Reason + Generativity + Research + Investigation. Science + Technology + Arts + History. (the last two are totally ignored by rationalists although they are of equal weight in the real world intellectual economy). Rationality is fragmented, specialised (incl. narrow AI) thinking. Creativity is unified, general thinking. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
Philosophically, "intelligence explosion" in the sense being discussed here is akin to ritual magic - the primary fallacy is the attribution to symbols alone of powers they simply do not possess. The argument is that an initially somewhat intelligent program A can generate a more intelligent program B, which in turn can generate... so on to Z. Let's stop and consider that first step, A to B. Clearly A cannot already have B encoded within itself, or the process is mere installation of already-existing software. So it must generate and evaluate candidates B1, B2 etc and choose the best one. On what basis does it choose? Most intelligent? But there's no such function as Intelligence(S) where S is a symbol system. There are functions F(S, E) where E is the environment, denoting the ability of S to produce useful results in that environment; "intelligence" is the word we use to refer to a family of such functions. So A must evaluate Bx in the context of the environment in which B is intended to operate. Furthermore, A can't evaluate by comparing Bx's answers in each potential situation to the correct ones - if A knew the correct answers in all situations, it would already be as intelligent as B. It has to work by feedback from the environment. If we step back and think about it, we really knew this already. In every case where humans, machines or biological systems exhibit anything that could be called an intelligence improvement - biological evolution, a child learning to talk, a scientific community improving its theories, engineers building better aeroplanes, programmers improving their software - it involves feedback from the environment. The mistake of trying to reach truth by pure armchair thought was understandable in ancient Greece. We now know better. So attractive as the image of a Transcendent Power popping out of a basement may be to us geeks, it doesn't have anything to do with reality. Making smarter machines in the real world is, like every other engineering activity, a process that has to take place _in_ the real world. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
[agi] Coin-flipping duplicates (was: Breaking Solomonoff induction (really))
On 6/23/08, Matt Mahoney <[EMAIL PROTECTED]> wrote: > --- On Sun, 6/22/08, Kaj Sotala <[EMAIL PROTECTED]> wrote: > > On 6/21/08, Matt Mahoney <[EMAIL PROTECTED]> wrote: > > > > > > Eliezer asked a similar question on SL4. If an agent > > flips a fair quantum coin and is copied 10 times if it > > comes up heads, what should be the agent's subjective > > probability that the coin will come up heads? By the > > anthropic principle, it should be 0.9. That is because if > > you repeat the experiment many times and you randomly > > sample one of the resulting agents, it is highly likely > > that will have seen heads about 90% of the time. > > > > That's the wrong answer, though (as I believe I pointed out when the > > question was asked over on SL4). The copying is just a red > > herring, it doesn't affect the probability at all. > > > > Since this question seems to confuse many people, I wrote a > > short Python program simulating it: > > http://www.saunalahti.fi/~tspro1/Random/copies.py > > The question was about subjective anticipation, not the actual outcome. It > depends on how the agent is programmed. If you extend your experiment so that > agents perform repeated, independent trials and remember the results, you > will find that on average agents will remember the coin coming up heads 99% > of the time. The agents have to reconcile this evidence with their knowledge > that the coin is fair. If the agent is rational, then its subjective anticipation should match the most likely outcome, no? Define "perform repeated, independent trials". That's a vague wording - I can come up with at least two different interpretations: a) Perform the experiment several times. If, on any of the trials, copies are created, then have all of them partake in the next trial as well, flipping a new coin and possibly being duplicated again (and quickly leading to an exponentially increasing number of copies). Carry out enough trials to eliminate the effect of random chance. Since every agent is flipping a fair coin each time, by the time you finish running the trials, all of them will remember seeing a roughly equal amount of heads and tails. Knowing this, a rational agent should anticipate this result, and not a 99% ratio. b) Perform the experiment several times. If, on any of the trials, copies are created, leave most of them be and only have one of them partake in the repeat trials. This will eventually result in a large number of copies who've most recently seen heads and at most one copy at a time who's most recently seen tails. But this doesn't tell us anything about the original question! The original situation was, "if you flip a coin and get copied on seeing heads, what result should you anticipate seeing", not "if you flip a coin several times, and on each time that heads turn up, copies of you get made and most are set aside while one keeps flipping the coin, should you anticipate eventually ending up in a group that has most recently seen heads". Yes, there is a high chance of ending up in such a group, but we again have a situation where the copying doesn't really affect things - this kind of wording is effectively the same as asking, "if you flip a coin and stop flipping once you see heads, should you on enough trials anticipate that the outcome you most recently saw was heads" - the copying only gives you a small chance to keep flipping anyway. The agent should still anticipate seeing an equal ratio of tails and heads beforehand, since that's what it will see, up to the point that it ends up in a position where it'll stop flipping the coin anymore. > It is a tricker question without multiple trials. The agent then needs to > model its own thought process (which is impossible for any Turing computable > agent to do with 100% accuracy). If the agent knows that it is programmed so > that if it observes an outcome R times out of N that it would expect the > probability to be R/N, then it would conclude "I know that I would observe > heads 99% of the time and therefore I would expect heads with probability > 0.99". But this programming would not make sense in a scenario with > conditional copying. That's right, it doesn't. > Here is an equivalent question. If you flip a fair quantum coin, and you are > killed with 99% probability conditional on the coin coming up tails, then, > when you look at the coin, what is your subjective anticipation of seeing > "heads"? What sense of equivalent do you mean? It isn't directly equivalent, since it will produce a somewhat different outcome on the single-trial (or repeated single trial) case. Previously all the possible outcomes would have either been in the "seen heads" or the "seen tails" category, this question adds the "hasn't seen anything, is dead" category. In the original experiment my expectation would have been 50:50 - here I have a 50% subjective anticipation of seeing "heads", a 0.5% anticipation of seeing "tails", and 49,5% an
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
On Mon, Jun 23, 2008 at 12:50 AM, William Pearson <[EMAIL PROTECTED]> wrote: > 2008/6/22 Vladimir Nesov <[EMAIL PROTECTED]>: > >> >> Two questions: >> 1) Do you know enough to estimate which scenario is more likely? > > Well since intelligence explosions haven't happened previously in our > light cone, it can't be a simple physical pattern, so I think > non-exploding intelligences have the evidence for being simpler on > their side. This message that I'm currently writing hasn't happened previously in out light code. By your argument, it is evidence for it being more difficult to write, than to recreate life on Earth and human intellect, which is clearly false, for all practical purposes. You should state that argument more carefully, in order for it to make sense. > So we might find them more easily. I also think I have > solid reasoning to think intelligence exploding is unlikely, which > requires paper length rather than post length. So it I think I do, but > should I trust my own rationality? But not too much, especially when the argument is not technical (which is clearly the case for questions such as this one). If argument is sound, you should be able to convince seed AI crowd too, even against their confirmation bias. If you can't convince them, then either they are idiots, or the argument is not good enough, which means that it's probably wrong, and so you yourself shouldn't place too high stakes on it. > Getting a bunch of people together to argue for both paths seems like > a good bet at the moment. Yes, if it will lead to a good estimation of which methodology is more likely to succeed. >> 2) What does this difference change for research at this stage? > > It changes the focus of research from looking for simple principles of > intelligence (that can be improved easily on the fly), to one that > expects intelligence creation to be a societal process over decades. > > It also makes secrecy no longer be the default position. If you take > the intelligence explosion scenario seriously you won't write anything > in public forums that might help other people make AI. As bad/ignorant > people might get hold of it and cause the first explosion. > I agree, but it works only if you know that the answer is correct, and (which you didn't address and which is critical for these issues) you won't build a doomsday machine as a result of your efforts, even if this particular path turns out to be more feasible. If you want to achieve artificial flight, you can start a research project that will try to figure out the fundamental principles of flying and will last a thousand years, or you can get a short cut, by climbing to a highest cliff in the world (which is no easy feat too), and jumping from it, thus achieving limited flying. Even if you have a good argument that cliff-climbing is a simpler technology than aerodynamics, choosing to climb is a wrong conclusion. -- Vladimir Nesov [EMAIL PROTECTED] http://causalityrelay.wordpress.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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
2008/6/23 Bob Mottram <[EMAIL PROTECTED]>: > 2008/6/22 William Pearson <[EMAIL PROTECTED]>: >> 2008/6/22 Vladimir Nesov <[EMAIL PROTECTED]>: >> Well since intelligence explosions haven't happened previously in our >> light cone, it can't be a simple physical pattern > > Probably the last "intelligence explosion" - a relatively rapid > increase in the degree of adaptability capabile of being exhibited by > an organism - was the appearance of the first Homo sapiens. The > number and variety of tools created by Homo sapiens compared to > earlier hominids indicate that this was one of the great leaps forward > in history (probably greatly facilitated by a more elaborate language > ability). I am using intelligence explosion to mean what would Eliezer mean by it. See http://www.overcomingbias.com/2008/06/optimization-an.html#more I.e. something never seen on this planet. I am sceptical of whether such a process is theoretically possible. 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Equivalent of the bulletin for atomic scientists or CRN for AI?
2008/6/22 William Pearson <[EMAIL PROTECTED]>: > 2008/6/22 Vladimir Nesov <[EMAIL PROTECTED]>: > Well since intelligence explosions haven't happened previously in our > light cone, it can't be a simple physical pattern Probably the last "intelligence explosion" - a relatively rapid increase in the degree of adaptability capabile of being exhibited by an organism - was the appearance of the first Homo sapiens. The number and variety of tools created by Homo sapiens compared to earlier hominids indicate that this was one of the great leaps forward in history (probably greatly facilitated by a more elaborate language ability). > If you take > the intelligence explosion scenario seriously you won't write anything > in public forums that might help other people make AI. As bad/ignorant > people might get hold of it and cause the first explosion. I don't fear intelligence, only ignorance. --- 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=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com