Re: [singularity] Scenarios for a simulated universe (second thought!)
Shane Legg wrote: :-) No offence taken, I was just curious to know what your position was. I can certainly understand people with a practical interest not having time for things like AIXI. Indeed as I've said before, my PhD is in AIXI and related stuff, and yet my own AGI project is based on other things. So even I am skeptical about whether it will lead to practical methods. That said, I can see that AIXI does have some fairly theoretical uses, perhaps Friendliness will turn out to be one of them? Shane ... As described (I haven't read, and probably couldn't read, the papers on AIXI, only on the list, and, when I get that far, in Ben's text) AIXI doesn't appear to be anything that a reasonable person would call intelligent. As such, I don't see how it could shed any light on Friendliness. Would you care to elaborate? Or were the descriptions on the list, perhaps, unfair? - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
[singularity] Scenarios for a simulated universe
i am familiar with 'simulation argument', various modes of philosophical/epistemological thinking about the nature of reality and simulation, and the previous replies to this mailing list. so am i prepared to share some brief words about the subject??? X-P do you ever get the sense that you are [merely] an instance of a (immortal information) template that describes patterns about your DNA (or physical hardware description) and thoughts (software)? in some moments, do you 'wake up' feeling like you've been re-started with a set of initial conditions (which don't necessarily need to relate to any 'real' past experience)? do you ever feel like one of 'The Sims' characters in a multi-dimensional simulation, that has video editing controls like play, pause, rewind, save, modify? that reality is not one monolithic continuous chain of events, but rather simulation fragments that do not mean anything in relation to each other aside from an extradimensional intelligence (ourselves?) that has artisitically designed them? can you understand the concept of determinism ( http://en.wikipedia.org/wiki/Determinism), the block universe model and virtual free will and no-separate-self ? how can one defend the position that this reality is NOT THE ONLY ONE? can every possible reality exists in its own multiverse possibility branch? Dr. Keta Meme http://ketameme.com [EMAIL PROTECTED] - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe (second thought!)
Ben, So you really think AIXI is totally useless? I haven't been reading Richard's comments, indeed I gave up reading his comments some time before he got himself banned from sl4, however it seems that you in principle support what he's saying. I just checked his posts and can see why they don't make sense, however I know very well that shouting rather than reasoning on the internet is a waste of time. My question to you then is a bit different. If you believe that AIXI is totally a waste of time, why is it that you recently published a book with a chapter on AIXI in it, and now think that AIXI and related study should be a significant part of what the SIAI does in the future? Shane - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Ben started a new thread about AIXI so I'll switch to there to keep this discussion in the same place and in sync with the subject line... Shane On 3/7/07, Mitchell Porter [EMAIL PROTECTED] wrote: From: Shane Legg [EMAIL PROTECTED] For sure. Indeed my recent paper on whether there exists an elegant theory of prediction tries to address that very problem. In short the paper says that if you want to convert something like Solomonoff induction or AIXI into a nice computable system... well you can't. Indeed my own work on building an intelligent machine is taking a neuro science inspired approach with just a few bits that are in some sense inspired by AIXI. I think the value of AIXI is that it gives you a relatively simple set of equations with which to mathematically study the properties of an ultra intelligent machine. In contrast something like Novamente can't be expressed in a one line equation. This makes it a much more difficult mathematical object to work with if you want to do theoretical analysis. This would be the paper, everyone: http://www.vetta.org/documents/IDSIA-12-06-1.pdf Shane - first you smack down the Goedel machine, and now AIXI! Is it genuinely useless in practice, do you think? Hutter says one of his current research priorities is to shrink it down into something that can run on existing machines... _ Advertisement: 50% off on Xbox 360, PS and Nintendo Wii titles! http://www.play-asia.com/SOap-23-83-4lab-71-bn-49-en-84-k-40-extended.html - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983 - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/7/07, Eugen Leitl [EMAIL PROTECTED] wrote: I realize that this is sarcasm, but detecting the mere presence of a species (nevermind their critical acclaim) from a trajectory, then rather give me the infinite simians, and I will personally look for Shakespeare sonnets in them. And I've seen equally vigorous handwaving in claims about the wondrous things AIXI on an infinite computer could do, so the analogy still holds nicely! - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Shane Legg recently (3-5-07) wrote: ...if you're not careful you may well define intelligence in such a way that humans don't have it either. I think it would be a serious mistake to degrade the definition of intelligence to the point that it included humans. Mike Deering, General Editor, http://nano-catalog.com/ Director, Singularity Action Group http://home.mchsi.com/~deering9/index.html Email: deering9 at mchsi dot com - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/5/07, John Ku [EMAIL PROTECTED] wrote: On 3/4/07, Ben Goertzel [EMAIL PROTECTED] wrote: Richard, I long ago proposed a working definition of intelligence as Achieving complex goals in complex environments. I then went through a bunch of trouble to precisely define all the component terms of that definition; you can consult the Appendix to my 2006 book The Hidden Pattern I'm not sure if your working definition is supposed to be significantly less ambitious than a philosophical definition or perhaps you even address something like this in your appendix, but I'm wondering whether the hypothetical example of Blockhead from philosophy of mind creates problems for your definition. Imagine that a computer has a huge memory bank of what actions to undertake given what inputs. With a big enough memory, it seems it could be perfectly capable of achieving complex goals in complex environments. Yet in doing so, there would be very little internal processing, just the bare minimum needed to look up and execute the part of its memory corresponding to its current inputs. I think any intuitive notion of intelligence would not count such a computer as being intelligent to any significant degree no matter how large its memory bank is or how complex and diverse an environment its memory allows it to navigate. There's simply too little internal processing going on for it to count as much more intelligent than any ordinary database application, though it might of course, do a pretty good job of fooling us into thinking it is intelligent if we don't know the details. I think this example actually poses a problem for any purely behavioristic definition of intelligence. To fit our ordinary notion of intelligence, I think there would have to be at least some sort of criteria concerning how the internal processing for the behavior is being done. I think the Blockhead example is normally presented in terms of looking up information from a huge memory bank, but as I'm thinking about it just now as I'm typing this up, I'm wondering if it could also be run with similar conclusions for simple brute search algorithms. If instead of a huge memory bank, it had enormous processing power and speed such that it could just explore every single chain of possibilities for the one that will lead to some specified goal, I'm not sure that would really count as intelligent to any significant degree either. You seem to be equating intelligence with consciousness. Ned Block also seems to do this in his original paper. I would prefer to reserve intelligence for third person observable behaviour, which would make the Blockhead intelligent, and consciousness for the internal state: it is possible that the Blockhead is unconscious or at least differently conscious compared to the human. Stathis Papaioannou - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/6/07, John Ku [EMAIL PROTECTED] wrote: On 3/5/07, Stathis Papaioannou [EMAIL PROTECTED] wrote: You seem to be equating intelligence with consciousness. Ned Block also seems to do this in his original paper. I would prefer to reserve intelligence for third person observable behaviour, which would make the Blockhead intelligent, and consciousness for the internal state: it is possible that the Blockhead is unconscious or at least differently conscious compared to the human. I think the argument also works for consciousness but I don't think you're right if you are suggesting that our ordinary notion of intelligence is merely third person observable behavior. (If you really were just voicing your own idiosyncratic preference for how you happen to like to use the term intelligence then I guess I don't really have a problem with that so long as you are clear about it.) Our ordinary notion of intelligence involves consciousness, but this term until relatively recently was taboo in cognitive science, the implication being that if it's not third person observable it doesn't exist, or at least we should pretend that it doesn't exist. It was against such a behaviourist view that the Blockhead argument was aimed. Stathis Papaioannou - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Ben, Would such an AIXI system have feelings or awareness? I have no idea, indeed I don't even know how to define such things outside of my own subject experience of them... Or to put it another way, if defining intelligence is hard, then defining some of these other things seems to be even harder. Shane - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: What I wanted was a set of non-circular definitions of such terms as intelligence and learning, so that you could somehow *demonstrate* that your mathematical idealization of these terms correspond with the real thing, ... so that we could believe that the mathematical idealizations were not just a fantasy. The last time I looked at a dictionary, all definitions are circular. So you win. Sigh! This is a waste of time: you just (facetiously) rejected the fundamental tenet of science. Which means that the stuff you were talking about was just pure mathematical fantasy, after all, and nothing to do with science, or the real world. Richard Loosemre. What does the definition of intelligence have to do with AIXI? AIXI is an optimization problem. The problem is to maximize an accumulated signal in an unknown environment. AIXI says the solution is to guess the simplest explanation for past observation (Occam's razor), and that this solution is not computable in general. I believe these principles have broad applicability to the design of machine learning algorithms, regardless of whether you consider such algorithms intelligent. You're going around in circles. If you were only talking about machine learning in the sense of an abstract mathematical formalism that has no relationship to learning, intelligence or anything going on in the real world, and in particular the real world in which some of us are interested in the problem of trying to build an intelligent system, then, fine, all power to you. At *that* level you are talking about a mathematical fantasy, not about science. But you did not do that: you made claims that went far beyond the confines of a pure, abstract mathematical formalism: you tried to relate that to an explanation of why Occam's Razor works (and remember, the original meaning of Occam's Razor was all about how an *intelligent* being should use its intelligence to best understand the world), and you also seemed to make inferences to the possibility that the real world was some kind of simulation. It seems to me that you are trying to have your cake and eat it too. Richard Loosemore. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
--- Richard Loosemore [EMAIL PROTECTED] wrote: Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: What I wanted was a set of non-circular definitions of such terms as intelligence and learning, so that you could somehow *demonstrate* that your mathematical idealization of these terms correspond with the real thing, ... so that we could believe that the mathematical idealizations were not just a fantasy. The last time I looked at a dictionary, all definitions are circular. So you win. Sigh! This is a waste of time: you just (facetiously) rejected the fundamental tenet of science. Which means that the stuff you were talking about was just pure mathematical fantasy, after all, and nothing to do with science, or the real world. Richard Loosemre. What does the definition of intelligence have to do with AIXI? AIXI is an optimization problem. The problem is to maximize an accumulated signal in an unknown environment. AIXI says the solution is to guess the simplest explanation for past observation (Occam's razor), and that this solution is not computable in general. I believe these principles have broad applicability to the design of machine learning algorithms, regardless of whether you consider such algorithms intelligent. You're going around in circles. If you were only talking about machine learning in the sense of an abstract mathematical formalism that has no relationship to learning, intelligence or anything going on in the real world, and in particular the real world in which some of us are interested in the problem of trying to build an intelligent system, then, fine, all power to you. At *that* level you are talking about a mathematical fantasy, not about science. But you did not do that: you made claims that went far beyond the confines of a pure, abstract mathematical formalism: you tried to relate that to an explanation of why Occam's Razor works (and remember, the original meaning of Occam's Razor was all about how an *intelligent* being should use its intelligence to best understand the world), and you also seemed to make inferences to the possibility that the real world was some kind of simulation. It seems to me that you are trying to have your cake and eat it too. I claim that AIXI has practical applications to machine learning. I also claim (implicitly) that machine learning has practical applications to the real world. Therefore, I claim that AIXI has practical applications to the real world (i.e. as Occam's razor). Further, because AIXI requires that the unknown environment be computable, I claim that we cannot exclude the possibility that the universe is a simulation. If Occam's razor did not work in practice, then you could claim that the universe is not computable, and therefore could not be a simulation. This really has nothing to do with the definition of intelligence. You can accept Turing's definition, which would exclude all animals except Homo Sapiens. You can accept a broader definition that would include machine learning. Both the human brain and linear regression algorithms make use of Occam's razor. I don't care if you call them intelligent or not. -- Matt Mahoney, [EMAIL PROTECTED] - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/5/07, Shane Legg [EMAIL PROTECTED] wrote: Would such an AIXI system have feelings or awareness? I have no idea, indeed I don't even know how to define such things outside of my own subject experience of them... I don't know how to define them either, but I can answer your question. What programs of the please run this on an infinite computer type (AIXI, Blockhead, a bunch of others with acronyms and cutesy names that I don't remember) actually amount to is suppose I am Jehovah, then I will create all possible universes [or all universes of a certain type] and select the ones with relevant properties. (Which is mathematically consistent though of no practical relevance seeing as one is not actually Jehovah.) In general, some of the universes so created will contain conscious beings. So a simple infinite computation would contain conscious beings, even though it would not itself be conscious. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Russell Wallace writes: What programs of the please run this on an infinite computer type (AIXI, Blockhead, a bunch of others with acronyms and cutesy names that I don't remember) actually amount to is suppose I am Jehovah, then I will create all possible universes [or all universes of a certain type] and select the ones with relevant properties. (Which is mathematically consistent though of no practical relevance seeing as one is not actually Jehovah.) It should be a fairly obvious implementation of a nested quantum computer to run any of these infinite processing programs. We will soon have oracle type computers that can answer any question with the reservation that the top level of the nest will have to be large enough to hold both the question and the answer. Current quantum computers are at the 3 or 4 bit level but scientists are confident of exponential advancement in future development. Of course then the problem will be, Are you smart enough to understand the answer? Mike Deering, General Editor, http://nano-catalog.com/ Director, Singularity Action Group http://home.mchsi.com/~deering9/index.html Email: deering9 at mchsi dot com - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Excellent point about questions. The question is the key to all of this, not knowledge or intelligence. Will these machines answer questions about existing knowledge or about knowledge that does not yet exist? Answering questions about existing knowledge is a search algorithm and enables learning. Automation of logic, e.g., can make this more powerful and make existing knowledge more accessible and usable, but it's still a search for some knowledge that already exists. Answering questions about knowledge that does not yet exist is knowledge creation and this is the process behind ALL social advance. Absolutely nothing intellectual advances without it. And there will be no singularity without understanding it. In fact, this understanding will be singularity. When we create knowledge today it occurs unconsciously or by accident. Most people spend their lives playing in a circular field of 'experts,' bantering about existing knowledge, which is often a huge waste of time and effort in terms of social advance. All of you just ask yourselves when you last created new knowledge. Not learned it, or read it, or argued it, or compiled it, or shared it, etc., but created it.What was it? How did you do it? Where did it come from? Where did it go? And if, by chance, you're not personally aware of when you did it, how could you possibly build a machine that is capable of doing it? Without KC, you're looking at something to make existing knoweldge more accessible, or interpretable, or practical/applicable, or learnable, etc., and you're leaving the rest up to the humans that by chance stumble upon this one process that moves everything intellectual forward. At best, building a helper of stumbling humans. Kind Regards, Bruce LaDuke Managing Director Instant Innovation, LLC Indianapolis, IN [EMAIL PROTECTED] http://www.hyperadvance.com Original Message Follows From: deering [EMAIL PROTECTED] Reply-To: singularity@v2.listbox.com To: singularity@v2.listbox.com Subject: Re: [singularity] Scenarios for a simulated universe Date: Mon, 5 Mar 2007 16:57:55 -0600 Russell Wallace writes: What programs of the please run this on an infinite computer type (AIXI, Blockhead, a bunch of others with acronyms and cutesy names that I don't remember) actually amount to is suppose I am Jehovah, then I will create all possible universes [or all universes of a certain type] and select the ones with relevant properties. (Which is mathematically consistent though of no practical relevance seeing as one is not actually Jehovah.) It should be a fairly obvious implementation of a nested quantum computer to run any of these infinite processing programs. We will soon have oracle type computers that can answer any question with the reservation that the top level of the nest will have to be large enough to hold both the question and the answer. Current quantum computers are at the 3 or 4 bit level but scientists are confident of exponential advancement in future development. Of course then the problem will be, Are you smart enough to understand the answer? Mike Deering, General Editor, http://nano-catalog.com/ Director, Singularity Action Group http://home.mchsi.com/~deering9/index.html Email: deering9 at mchsi dot com - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983 _ Dont miss your chance to WIN 10 hours of private jet travel from Microsoft® Office Live http://clk.atdmt.com/MRT/go/mcrssaub0540002499mrt/direct/01/ - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
From: deering [EMAIL PROTECTED] It should be a fairly obvious implementation of a nested quantum computer to run any of these infinite processing programs. We will soon have oracle type computers that can answer any question with the reservation that the top level of the nest will have to be large enough to hold both the question and the answer. Current quantum computers are at the 3 or 4 bit level but scientists are confident of exponential advancement in future development. Of course then the problem will be, Are you smart enough to understand the answer? You radically overstate the expected capabilities of quantum computers. They can't even do NP-complete problems in polynomial time. http://scottaaronson.com/blog/?p=208 _ Advertisement: Fresh jobs daily. Stop waiting for the newspaper. Search Now! www.seek.com.au http://a.ninemsn.com.au/b.aspx?URL=http%3A%2F%2Fninemsn%2Eseek%2Ecom%2Eau_t=757263760_r=Hotmail_EndText_Dec06_m=EXT - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/6/07, Mitchell Porter [EMAIL PROTECTED] wrote: You radically overstate the expected capabilities of quantum computers. They can't even do NP-complete problems in polynomial time. http://scottaaronson.com/blog/?p=208 What about a computer (classical will do) granted an infinity of cycles through, for example, a Freeman Dyson or Frank Tipler type mechanism? No matter how many cycles it takes to compute a particular simulated world, any delay will be transparent to observers in that world. It only matters that the computation doesn't stop before it is completed. Stathis Papaioannou - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
--- Stathis Papaioannou [EMAIL PROTECTED] wrote: On 3/6/07, Mitchell Porter [EMAIL PROTECTED] wrote: You radically overstate the expected capabilities of quantum computers. They can't even do NP-complete problems in polynomial time. http://scottaaronson.com/blog/?p=208 What about a computer (classical will do) granted an infinity of cycles through, for example, a Freeman Dyson or Frank Tipler type mechanism? No matter how many cycles it takes to compute a particular simulated world, any delay will be transparent to observers in that world. It only matters that the computation doesn't stop before it is completed. The computation would also require infinite memory (a Turing machine), or else it would cycle. Although our universe might be the product of a Turing machine, the physics of our known universe will only allow finite memory. The number of possible quantum states of a closed system with finite size and mass is finite. For our universe (big bang model), the largest memory you could construct would be on the order of c^5 T^2/hG ~ 10^122 bits (where c is the speed of light, T is the age of the universe, h is Planck's constant and G is the gravitational constant. (Coincidentally, each bit would occupy about the volume of a proton or neutron). A quantum computer is weaker than a finite state machine. A quantum computer is restricted to time-reversible computation, so operations like bit assignment or copying are not allowed. And even if you had a Turing machine, you still could not compute a solution to AIXI. It is not computable, like the halting problem. -- Matt Mahoney, [EMAIL PROTECTED] - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: What I wanted was a set of non-circular definitions of such terms as intelligence and learning, so that you could somehow *demonstrate* that your mathematical idealization of these terms correspond with the real thing, ... so that we could believe that the mathematical idealizations were not just a fantasy. The last time I looked at a dictionary, all definitions are circular. So you win. Sigh! This is a waste of time: you just (facetiously) rejected the fundamental tenet of science. Which means that the stuff you were talking about was just pure mathematical fantasy, after all, and nothing to do with science, or the real world. Richard Loosemre. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Ben Goertzel wrote: Richard Loosemore wrote: Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: What I wanted was a set of non-circular definitions of such terms as intelligence and learning, so that you could somehow *demonstrate* that your mathematical idealization of these terms correspond with the real thing, ... so that we could believe that the mathematical idealizations were not just a fantasy. The last time I looked at a dictionary, all definitions are circular. So you win. Richard, I long ago proposed a working definition of intelligence as Achieving complex goals in complex environments. I then went through a bunch of trouble to precisely define all the component terms of that definition; you can consult the Appendix to my 2006 book The Hidden Pattern Shane Legg and Marcus Hutter have proposed a related definition of intelligence in a recent paper... Anyone can propose a definition. The point of my objection is that a definition has to have some way to be compared against reality. Suppose I define intelligence to be: A funtion that maps goals G and world states W onto action states A, where G, W and A are any mathematical entities whatsoever. That would make any function that maps X [cross] Y into Z an intelligence. Such a definition would be pointless. The question is *why* would it be pointless? What criteria are applied, in order to determine whether the definition has something to the thing that in everyday life we call intelligence. My protest to Matt was that I did not believe his definition could be made to lead to anything like a reasonable grounding. I tried to get him to do the grounding, but to no avail: he eventually resorted to the blanket denial that any definition means anything ... which is a cop out if he wanted to defend the claim that the formalism was something more than a mathematical fantasy. Richard Loosemore P.S. Quick sanity check: you know the last comment in the quote you gave (about loking in the dictionary) was Matt's, not mine, right? - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Richard, I long ago proposed a working definition of intelligence as Achieving complex goals in complex environments. I then went through a bunch of trouble to precisely define all the component terms of that definition; you can consult the Appendix to my 2006 book The Hidden PatternShane Legg and Marcus Hutter have proposed a related definition of intelligence in a recent paper... Anyone can propose a definition. The point of my objection is that a definition has to have some way to be compared against reality. Suppose I define intelligence to be: A funtion that maps goals G and world states W onto action states A, where G, W and A are any mathematical entities whatsoever. That would make any function that maps X [cross] Y into Z an intelligence. Such a definition would be pointless. The question is *why* would it be pointless? What criteria are applied, in order to determine whether the definition has something to the thing that in everyday life we call intelligence. The difficulty in comparing my definition against reality is that my definition defines intelligence relative to a complexity measure. For this reason, it is fundamentally a subjective definition of intelligence, except in the unrealistic case where degree of complexity tends to infinity (in which case all reasonably general complexity measures become equivalent, due to bisimulation of Turing machines). To qualitatively compare my definition to the everyday life definition of intelligence, we can check its consistency with our everyday life definition of complexity. Informally, at least, my definition seems to check out to me: intelligence according to an IQ test does seem to have something to do with the ability to achieve complex goals; and, the reason we think IQ tests mean anything is that we think the ability to achieve complex goals in the test-context will correlate with the ability to achieve complex goals in various more complex environments (contexts). Anyway, if I accept for instance **Richard Loosemore** as a measurer of the complexity of environments and goals, then relative to Richard-as-a-complexity-measure, I can assess the intelligence of various entities, using my definition In practice, in building a system like Novamente, I'm relying on modern human culture's consensus complexity measure and trying to make a system that, according to this measure, can achieve a diverse variety of complex goals in complex situations... P.S. Quick sanity check: you know the last comment in the quote you gave (about loking in the dictionary) was Matt's, not mine, right? Yes... Ben - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Definition is intelligence. Kind Regards, Bruce LaDuke Managing Director Instant Innovation, LLC Indianapolis, IN [EMAIL PROTECTED] http://www.hyperadvance.com Original Message Follows From: Ben Goertzel [EMAIL PROTECTED] Reply-To: singularity@v2.listbox.com To: singularity@v2.listbox.com Subject: Re: [singularity] Scenarios for a simulated universe Date: Sun, 04 Mar 2007 14:26:33 -0500 Richard, I long ago proposed a working definition of intelligence as Achieving complex goals in complex environments. I then went through a bunch of trouble to precisely define all the component terms of that definition; you can consult the Appendix to my 2006 book The Hidden PatternShane Legg and Marcus Hutter have proposed a related definition of intelligence in a recent paper... Anyone can propose a definition. The point of my objection is that a definition has to have some way to be compared against reality. Suppose I define intelligence to be: A funtion that maps goals G and world states W onto action states A, where G, W and A are any mathematical entities whatsoever. That would make any function that maps X [cross] Y into Z an intelligence. Such a definition would be pointless. The question is *why* would it be pointless? What criteria are applied, in order to determine whether the definition has something to the thing that in everyday life we call intelligence. The difficulty in comparing my definition against reality is that my definition defines intelligence relative to a complexity measure. For this reason, it is fundamentally a subjective definition of intelligence, except in the unrealistic case where degree of complexity tends to infinity (in which case all reasonably general complexity measures become equivalent, due to bisimulation of Turing machines). To qualitatively compare my definition to the everyday life definition of intelligence, we can check its consistency with our everyday life definition of complexity. Informally, at least, my definition seems to check out to me: intelligence according to an IQ test does seem to have something to do with the ability to achieve complex goals; and, the reason we think IQ tests mean anything is that we think the ability to achieve complex goals in the test-context will correlate with the ability to achieve complex goals in various more complex environments (contexts). Anyway, if I accept for instance **Richard Loosemore** as a measurer of the complexity of environments and goals, then relative to Richard-as-a-complexity-measure, I can assess the intelligence of various entities, using my definition In practice, in building a system like Novamente, I'm relying on modern human culture's consensus complexity measure and trying to make a system that, according to this measure, can achieve a diverse variety of complex goals in complex situations... P.S. Quick sanity check: you know the last comment in the quote you gave (about loking in the dictionary) was Matt's, not mine, right? Yes... Ben - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983 _ Play Flexicon: the crossword game that feeds your brain. PLAY now for FREE. http://zone.msn.com/en/flexicon/default.htm?icid=flexicon_hmtagline - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
--- Richard Loosemore [EMAIL PROTECTED] wrote: Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: What I wanted was a set of non-circular definitions of such terms as intelligence and learning, so that you could somehow *demonstrate* that your mathematical idealization of these terms correspond with the real thing, ... so that we could believe that the mathematical idealizations were not just a fantasy. The last time I looked at a dictionary, all definitions are circular. So you win. Sigh! This is a waste of time: you just (facetiously) rejected the fundamental tenet of science. Which means that the stuff you were talking about was just pure mathematical fantasy, after all, and nothing to do with science, or the real world. Richard Loosemre. What does the definition of intelligence have to do with AIXI? AIXI is an optimization problem. The problem is to maximize an accumulated signal in an unknown environment. AIXI says the solution is to guess the simplest explanation for past observation (Occam's razor), and that this solution is not computable in general. I believe these principles have broad applicability to the design of machine learning algorithms, regardless of whether you consider such algorithms intelligent. -- Matt Mahoney, [EMAIL PROTECTED] - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/4/07, Matt Mahoney wrote: What does the definition of intelligence have to do with AIXI? AIXI is an optimization problem. The problem is to maximize an accumulated signal in an unknown environment. AIXI says the solution is to guess the simplest explanation for past observation (Occam's razor), and that this solution is not computable in general. I believe these principles have broad applicability to the design of machine learning algorithms, regardless of whether you consider such algorithms intelligent. Matt, you might want to consider that while Occam's Razor is indeed a very beautiful and powerful principle, it is a heuristic directly applicable only to those situations of all else being equal (or made effectively so by means of infinite computing power.) [Observant readers may notice than I'm being slightly tongue in cheek here, drawing a parallel with a recent mismatch of expressed views on the AGI and Extropy lists regarding the elegance of the Principle of Indifference. The analogy is sublime.] My point is that nature never directly applies the perfect principle. Every problem posed to nature carries an implicit bias, and this is enough to start nature down the path toward a satisficing heuristic. While the Principle of Parsimony and the Principle of Indifference play unattainably objective roles in our epistemology, you may want to consider their subjective cousin, Max Entropy, as one of your star players in any practical AI. - Jef - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
--- Richard Loosemore [EMAIL PROTECTED] wrote: What I wanted was a set of non-circular definitions of such terms as intelligence and learning, so that you could somehow *demonstrate* that your mathematical idealization of these terms correspond with the real thing, ... so that we could believe that the mathematical idealizations were not just a fantasy. The last time I looked at a dictionary, all definitions are circular. So you win. P.S. The above definition is broken anyway: what about unsupervised learning? What about learning by analogy? I should have specified supervised learning as an application of AIXI. There are subsets, H, of Turing machines for which there are efficient algorithms for finding a small h in H that is consistent with the training data. Examples include decision trees, neural networks, polynomial regression, clustering, etc. However AIXI does not necessarily imply learning. There are other approaches. -- Matt Mahoney, [EMAIL PROTECTED] - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
--- Ben Goertzel [EMAIL PROTECTED] wrote: Matt, I really don't see why you think Hutter's work shows that Occam's Razor holds in any context except AI's with unrealistically massive amounts of computing power (like AIXI and AIXItl) In fact I think that it **does** hold in other contexts (as a strategy for reasoning by modest-resources minds like humans or Novamente), but I don't see how Hutter's work shows this... I admit Hutter did not make claims about machine learning frameworks or Occam's razor, but we should not view his work in such narrow context. Hutter's conclusions about the optimal behavior of rational agents were proven for the following cases: 1. Unrestricted environments (in which case the solution is not computable), 2. Space and time bounded environments (in which case the solution is intractable), 3. Subsets of (1) or (2) such that the environment is consistent with past interaction. But the same reasoning he used in his proofs could just as well be applied to practical cases of machine learning for which efficient solutions are known. The proofs all use the fact that shorter Turing machines are more likely than longer ones (a Solomonoff prior). For example, Hutter does not tell us how to solve linear regression, fitting a straight line to a set of points. What Hutter tells us is two other things: 1. Linear regression is a good predictor, even though a higher order polynomial might have a better fit (because a low order polynomial has lower algorithmic complexity). 2. Linear regression is useful, even though other machine learning algorithms might be better predictors (because a general solution is not computable, so we have to settle for a suboptimal solution). So what I did was two things. First, I used the fact that Occam's razor works in both simulated and real environments (based on extensions of AIXI and empirical observations respectively) to argue that the universe is consistent with a simulation. (This is disturbing because you are not programmed to think this way). Second, I used the same reasoning to guess about the nature of the universe (assuming it is simulated), and the only thing we know is that shorter simulation programs are more likely than longer ones. My conclusion was that bizarre behavior or a sudden end is unlikely, because such events would not occur in the simplest programs. This ought to at least be reassuring. -- Matt Mahoney -- Matt Mahoney, [EMAIL PROTECTED] - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Matt, When you said (in the text below): In every practical case of machine learning, whether it is with decision trees, neural networks, genetic algorithms, linear regression, clustering, or whatever, the problem is you are given training pairs (x,y) and you have to choose a hypothesis h from a hypothesis space H that best classifies novel test instances, h(x) = y. ... you did *exactly* what I was complaining about. Correct me if I am wrong, but it looks like you just declared learning to be a particular class of mathematical optimization problem, without making reference to the fact that there is a more general meaning of learning that is vastly more complex than your above definition. What I wanted was a set of non-circular definitions of such terms as intelligence and learning, so that you could somehow *demonstrate* that your mathematical idealization of these terms correspond with the real thing, ... so that we could believe that the mathematical idealizations were not just a fantasy. If what you gave was supposed to be a definition, then it was circular (you defined learning to *be* the idealization). The rest of what you say (about Occam's Razor etc.) is irrelevant if you or Hutter cannot prove something more than a hand-waving connection between the mathematical idealizations of intelligence, learning, etc., and the original meanings of those words. So my original request stands unanswered. Richard Loosemore. P.S. The above definition is broken anyway: what about unsupervised learning? What about learning by analogy? Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: Matt Mahoney wrote: --- Richard Loosemore [EMAIL PROTECTED] wrote: Matt Mahoney wrote: As you probably know, Hutter proved that the optimal behavior of a goal seeking agent in an unknown environment (modeled as a pair of interacting Turing machines, with the enviroment sending an additional reward signal to the agent that the agent seeks to maximize) is for the agent to guess at each step that the environment is modeled by the shortest program consistent with the observed interaction so far. The proof requires the assumption that the environment be computable. Essentially, the proof says that Occam's Razor is the best general strategy for problem solving. The fact that this works in practice strongly suggests that the universe is indeed a simulation. It suggests nothing of the sort. Hutter's theory is a mathematical fantasy with no relationship to the real world. Hutter's theory makes a very general statement about the optimal behavior of rational agents. Is this really irrelevant to the field of machine learning? Define rational agent. Define optimal behavior. In the framework of Hutter's AIXI, optimal behavior is the behavior that maximizes the accumulated reward signal from the environment. In general, this problem is not computable. (It is equivalent to solving the Kolmogorov complexity of the environment). An agent with limited computational resources is rational if it chooses the best strategy within those limits for maximizing its accumulated reward signal (in general, a suboptimal solution). Then prove that a rational agent following optimal behavior is actually intelligent (as we in colloquial speech use the word intelligent), and do this *without* circularly defining the meaning of intelligence to be, in effect, the optimal behavior of a rational agent. Turing defined an agent as intelligent if communication with it is indistinguishable from human. This is not the same as rational behavior, but it is probably the best definition we have. One caveat: Don't come back and ask me to be precise about what we in colloquial speech mean when we use the word intelligent, because some of us who reject this theory would state that the term does not have an analytic definition, only an empirical one. Your position, on the other hand, is that a precise definition does exist and that you know what it is when you say that a rational agent following optimal behavior is an intelligent system. For this reason the onus is on you (and not me) to say what intelligence is. My claim is that you cannot, without circularity, prove that rational agents following optimal behavior are the same thing as intelligent systems, and for that reason your use of all of these terms is just unsubstantiated speculation. Labels attached to an abstract mathematical formalism with nothing but your intuition in the way of justification. This unsubstantiated speculation then escalates into a zone of complete nonsense when it talks about hypothetical systems of infinite size and power, without showing in any way why we should believe that the properties of such infinitely large systems carry over to systems in the real world. Hence, it is a mathematical fantasy with no relationship to the real world. QED. Richard Loosemore. Hutter realizes
Re: [singularity] Scenarios for a simulated universe
On 3/2/07, Matt Mahoney [EMAIL PROTECTED] wrote: Second, I used the same reasoning to guess about the nature of the universe (assuming it is simulated), and the only thing we know is that shorter simulation programs are more likely than longer ones. My conclusion was that bizarre behavior or a sudden end is unlikely, because such events would not occur in the simplest programs. This ought to at least be reassuring. Consider that while the trunk of the universal tree of the probable grows increasingly stable, the branches do often swing in the winds, and many of the thinner branches of the possible do not survive. Do you assume that humanity is presently nestled in the crook of a highly probable branch? If our own branch were to break, would you take comfort in knowing that the tree itself stands strong? - Jef - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/1/07, Matt Mahoney [EMAIL PROTECTED] wrote: As you probably know, Hutter proved that the optimal behavior of a goal seeking agent in an unknown environment (modeled as a pair of interacting Turing machines, with the enviroment sending an additional reward signal to the agent that the agent seeks to maximize) is for the agent to guess at each step that the environment is modeled by the shortest program consistent with the observed interaction so far. The proof requires the assumption that the environment be computable. Essentially, the proof says that Occam's Razor is the best general strategy for problem solving. The fact that this works in practice strongly suggests that the universe is indeed a simulation. With this in mind, I offer 5 possible scenarios ranked from least to most likely based on the Kolmogorov complexity of the simulator. I think this will allay any fears that our familiar universe might suddenly be switched off or behave in some radically different way. 1. Neurological level. Your brain is connected to a computer at all the input and output points, e.g. the spinal cord, optic and auditory nerves, etc. The simulation presents the illusion of a human body and a universe containing billions of other people like yourself (but not exactly alike). The algorithmic complexity of this simulation would be of the same order as the complexity of your brain, about 10^13 bits (by counting synapses). 2. Cognitive level. Rather than simulate the entire brain, the simulation includes all of the low level sensorimotor processing as part of the environment. For example, when you walk you don't think about the contraction of individual leg muscles. When you read this, you think about the words and not the arrangement of pixels in your visual field. That type of processing is part of the environment. You are presented with a universe at the symbolic level of words and high-level descriptions. This is about 10^9 bits, based on the amount of verbal information you process in a lifetime, and estimates of long term memory capacity by Standing and Landauer. 3. Biological level. Unlike 1 and 2, you are not the sole intelligent being in the universe, but there is no life beyond Earth. The environment is a model of the Earth with just enough detail to simulate reality. Humans are modeled at the biological level. The complexity of a human model is that of our DNA. I estimate 10^7 bits. I know the genome is 6 x 10^9 bits uncompressed, but only about 2% of our DNA is biologically active. Also, many genes are copied many times, and there are equivalent codons for the same amino acids, genes can be moved and reordered, etc. 4. Physical level. A program simulates the fundamental laws of physics, with the laws tuned to allows life to evolve, perhaps on millions of planets. For example, the ratio of the masses of the proton and neutron is selected to allow the distribution of elements like carbon and oxygen needed for life to evolve. (If the neutron were slightly heavier, there would be no hydrogen fusion in stars. If it were slightly lighter, the proton would be unstable and all matter would decay into neutron bodies.) Likewise the force of gravity is set just right to allow matter to condense into stars and planets and not all collapse into black holes. Wolfram estimates that the physical universe can be modeled with just a few lines of code (see http://en.wikipedia.org/wiki/A_New_Kind_of_Science ), on the order of hundreds of bits. This is comparable to the information needed to set the free parameters of some string theories. 5. Mathematical level. The universe we observe is one of an enumeration of all Turing machines. Some universes will support life and some won't. We must, of course, be in one that will. The simulation is simply expressed as N, the set of natural numbers. Each level increases the computational requirements, while decreasing the complexity of the program and making the universe more predictable. You don't need much of a computer for level 5. A single physical state, perhaps the null state, can be considered an infinitely parallel computer mapping onto the natural numbers - indeed, mapping onto any computation you like under the right interpretation. This is sort of trivially obvious, like the assertion that a short string of symbols contains every possible book in every possible language if you interpret and re-interpret the symbols in the right way. In the case of the string, this isn't very interesting because you need to have the book before you can find the book. But in the case of computations, those which have observers will, as you suggest, self-select. Stathis Papaioannou - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
Matt Mahoney wrote: As you probably know, Hutter proved that the optimal behavior of a goal seeking agent in an unknown environment (modeled as a pair of interacting Turing machines, with the enviroment sending an additional reward signal to the agent that the agent seeks to maximize) is for the agent to guess at each step that the environment is modeled by the shortest program consistent with the observed interaction so far. The proof requires the assumption that the environment be computable. Essentially, the proof says that Occam's Razor is the best general strategy for problem solving. The fact that this works in practice strongly suggests that the universe is indeed a simulation. It suggests nothing of the sort. Hutter's theory is a mathematical fantasy with no relationship to the real world. Richard Loosemore. - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
--- Richard Loosemore [EMAIL PROTECTED] wrote: Matt Mahoney wrote: As you probably know, Hutter proved that the optimal behavior of a goal seeking agent in an unknown environment (modeled as a pair of interacting Turing machines, with the enviroment sending an additional reward signal to the agent that the agent seeks to maximize) is for the agent to guess at each step that the environment is modeled by the shortest program consistent with the observed interaction so far. The proof requires the assumption that the environment be computable. Essentially, the proof says that Occam's Razor is the best general strategy for problem solving. The fact that this works in practice strongly suggests that the universe is indeed a simulation. It suggests nothing of the sort. Hutter's theory is a mathematical fantasy with no relationship to the real world. Hutter's theory makes a very general statement about the optimal behavior of rational agents. Is this really irrelevant to the field of machine learning? As for whether the universe is real or simulated, nobody can prove one way or the other. But your brain is programmed through evolution to believe the universe is real. If you were programming an autonomous agent for self survival, wouldn't you program it that way? -- Matt Mahoney, [EMAIL PROTECTED] - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/1/07, Matt Mahoney [EMAIL PROTECTED] wrote: What I argue is this: the fact that Occam's Razor holds suggests that the universe is a computation. Matt - Would you please clarify how/why you think B follows from A in your preceding statement? - Jef - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/1/07, Matt Mahoney [EMAIL PROTECTED] wrote: --- Jef Allbright [EMAIL PROTECTED] wrote: On 3/1/07, Matt Mahoney [EMAIL PROTECTED] wrote: What I argue is this: the fact that Occam's Razor holds suggests that the universe is a computation. Matt - Would you please clarify how/why you think B follows from A in your preceding statement? Hutter's proof requires that the environment have a computable distribution. http://www.hutter1.net/ai/aixigentle.htm So in any universe of this type, Occam's Razor should hold. If Occam's Razor did not hold, then we could conclude that the universe is not computable. The fact that Occam's Razor does hold means we cannot rule out the possibility that the universe is simulated. Matt - I think this answers my question to you, at least I think I see where you're coming from. I would say that you have justification for saying that interaction with the universe demonstrates mathematically modelable regularities (in keeping with the principle of parsimony), rather than saying that it's a simulation (which involves additional assumptions.) Do you think you have information to warrant taking it further? - Jef - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
--- Jef Allbright [EMAIL PROTECTED] wrote: On 3/1/07, Matt Mahoney [EMAIL PROTECTED] wrote: --- Jef Allbright [EMAIL PROTECTED] wrote: On 3/1/07, Matt Mahoney [EMAIL PROTECTED] wrote: What I argue is this: the fact that Occam's Razor holds suggests that the universe is a computation. Matt - Would you please clarify how/why you think B follows from A in your preceding statement? Hutter's proof requires that the environment have a computable distribution. http://www.hutter1.net/ai/aixigentle.htm So in any universe of this type, Occam's Razor should hold. If Occam's Razor did not hold, then we could conclude that the universe is not computable. The fact that Occam's Razor does hold means we cannot rule out the possibility that the universe is simulated. Matt - I think this answers my question to you, at least I think I see where you're coming from. I would say that you have justification for saying that interaction with the universe demonstrates mathematically modelable regularities (in keeping with the principle of parsimony), rather than saying that it's a simulation (which involves additional assumptions.) Do you think you have information to warrant taking it further? - Jef There is no way to know if the universe is real or simulated. From our point of view, there is no difference. If the simulation is realistic then there is no experiment we could do to make the distinction. I am just saying that our universe is consistent with a simulation in that it appears to be computable. One disturbing implication is that the simulation might be suddenly turned off or changed in some radical way you can't anticipate. You really don't know anything about the world in which the simulation is being run. (The movie The Matrix is based on this idea). Maybe the Singularity has already happened and what you observe as the universe is part of the resulting computation. My argument is that if the universe is simulated then these possibilities are unlikely. My reasoning is that if we know nothing about this computation then we should assume a universal Solomonoff prior, i.e. a universal Turing machine programmed by random coin flips. This is what Hutter did to solve the problem of rational agents. I am applying the idea to understanding a universe about which (if it is not real) we know nothing, except that shorter programs are more likely than longer ones. -- Matt Mahoney, [EMAIL PROTECTED] - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
Re: [singularity] Scenarios for a simulated universe
On 3/1/07, Matt Mahoney [EMAIL PROTECTED] wrote: --- Jef Allbright [EMAIL PROTECTED] wrote: Matt - I think this answers my question to you, at least I think I see where you're coming from. I would say that you have justification for saying that interaction with the universe demonstrates mathematically modelable regularities (in keeping with the principle of parsimony), rather than saying that it's a simulation (which involves additional assumptions.) Do you think you have information to warrant taking it further? - Jef There is no way to know if the universe is real or simulated. From our point of view, there is no difference. If the simulation is realistic then there is no experiment we could do to make the distinction. I think you mean if the simulation is consistent then there's no experiment we could do to make the distinction. I am just saying that our universe is consistent with a simulation in that it appears to be computable. I agree with you that it seems there's nothing more we can say in that case about whether or not it's a simulation. One disturbing implication is that the simulation might be suddenly turned off or changed in some radical way you can't anticipate. Hmm, I thought you just made the perfectly good point that there's nothing further we can say about whether or not our world is a simulation, so what basis do you have for worrying about whether the simulation might be turned off? In fact, I have it on good faith that it was in fact turned off, for about 3M real years, just around breakfast time this morning. It'll probably be shut down again, but what could possibly be disturbing about something that can't possibly be detected? - Jef - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983
[singularity] Scenarios for a simulated universe
As you probably know, Hutter proved that the optimal behavior of a goal seeking agent in an unknown environment (modeled as a pair of interacting Turing machines, with the enviroment sending an additional reward signal to the agent that the agent seeks to maximize) is for the agent to guess at each step that the environment is modeled by the shortest program consistent with the observed interaction so far. The proof requires the assumption that the environment be computable. Essentially, the proof says that Occam's Razor is the best general strategy for problem solving. The fact that this works in practice strongly suggests that the universe is indeed a simulation. With this in mind, I offer 5 possible scenarios ranked from least to most likely based on the Kolmogorov complexity of the simulator. I think this will allay any fears that our familiar universe might suddenly be switched off or behave in some radically different way. 1. Neurological level. Your brain is connected to a computer at all the input and output points, e.g. the spinal cord, optic and auditory nerves, etc. The simulation presents the illusion of a human body and a universe containing billions of other people like yourself (but not exactly alike). The algorithmic complexity of this simulation would be of the same order as the complexity of your brain, about 10^13 bits (by counting synapses). 2. Cognitive level. Rather than simulate the entire brain, the simulation includes all of the low level sensorimotor processing as part of the environment. For example, when you walk you don't think about the contraction of individual leg muscles. When you read this, you think about the words and not the arrangement of pixels in your visual field. That type of processing is part of the environment. You are presented with a universe at the symbolic level of words and high-level descriptions. This is about 10^9 bits, based on the amount of verbal information you process in a lifetime, and estimates of long term memory capacity by Standing and Landauer. 3. Biological level. Unlike 1 and 2, you are not the sole intelligent being in the universe, but there is no life beyond Earth. The environment is a model of the Earth with just enough detail to simulate reality. Humans are modeled at the biological level. The complexity of a human model is that of our DNA. I estimate 10^7 bits. I know the genome is 6 x 10^9 bits uncompressed, but only about 2% of our DNA is biologically active. Also, many genes are copied many times, and there are equivalent codons for the same amino acids, genes can be moved and reordered, etc. 4. Physical level. A program simulates the fundamental laws of physics, with the laws tuned to allows life to evolve, perhaps on millions of planets. For example, the ratio of the masses of the proton and neutron is selected to allow the distribution of elements like carbon and oxygen needed for life to evolve. (If the neutron were slightly heavier, there would be no hydrogen fusion in stars. If it were slightly lighter, the proton would be unstable and all matter would decay into neutron bodies.) Likewise the force of gravity is set just right to allow matter to condense into stars and planets and not all collapse into black holes. Wolfram estimates that the physical universe can be modeled with just a few lines of code (see http://en.wikipedia.org/wiki/A_New_Kind_of_Science ), on the order of hundreds of bits. This is comparable to the information needed to set the free parameters of some string theories. 5. Mathematical level. The universe we observe is one of an enumeration of all Turing machines. Some universes will support life and some won't. We must, of course, be in one that will. The simulation is simply expressed as N, the set of natural numbers. Each level increases the computational requirements, while decreasing the complexity of the program and making the universe more predictable. -- Matt Mahoney, [EMAIL PROTECTED] - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983