Re: [singularity] Apology to the list, and a more serious commentary on AIXI
I'm still relatively new to the field so of course i do not put a lot of strength behind my opinions. I think there are some clarifications here that could be useful, although it is highly improbable that you need them. My overall argument is completely vindicated by what you say here. (My wording was sometimes ambiguous in that last email, I confess, but what I have been targeting is AIXI as proof, not AIXI as actual working system). I only care about where AIXI gets the power of its proof, so it does not matter to me whether a practical implementation [sic] of AIXI would actually need to build a cognitive system. It is not important whether it would do so in practice, because if the proof says that AIXI is allowed to build a complete cognitive system in the course of solving the IQ test problem, then what is the meaning of AIXI would equal any other intelligence starting with the same initial knowledge set well, yeah, of course it would, if it was allowed to build something as sophisticated as that other intelligence! It is like me saying I can prove that I can make a jet airliner with my bare hands . and then when you delve into the proof you find that my definition of make includes the act of putting in a phone call to Boeing and asking them to deliver one. Such a proof is completely valueless. It is rather vacuous that AIXI's capability to (to some degree) simulate another agent's [intellect] means it is on par with that agent['s intellect]. However, AIXI also proves the antecedent and it is not merely an assumption. This is different from s ilar circular arguments (in disguise). On top of this, it proves that we can do better than this intellect and that we may do optimal given limited data (in terms of what can be done given the data). I think your example is a strawman argument here. What is counter-intuitive is the definition of 'make' I believe. If you allow this to fall within the definition, and take care to formulate a strict definition and not one of varrying strength then any path is of equal worth. Not everything is able to phone Boeing and it is therefor not valueluess to know it is able to - it does contribute information and it is merely a choice of scale. What would you say that in your process of building the plane, your observation of birds and the wind led you to think in similar manners, and not nec. exactly the same, as the Wright's? If we do not allow a being to even implicitely ( i.e. without nec. realizing that it merely mimics another agent's train of thought, albeit I do not personally consider it relevant), or having the capabilities, to simulate that of another agent, in the world where every possible agent does (or has) existed, any additional agents are condemned. AIXI is valueless. QED. 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 - 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] Apology to the list, and a more serious commentary on AIXI
Ben Goertzel wrote: Sorry, but I simply do not accept that you can make do really well on a long series of IQ tests into a computable function without getting tangled up in an implicit homuncular trap (i.e. accidentally assuming some real intelligence in the computable function). Let me put it this way: would AIXI, in building an implementation of this function, have to make use of a universe (or universe simulation) that *implicitly* included intelligences that were capable of creating the IQ tests? So, if there were a question like this in the IQ tests: Anna Nicole is to Monica Lewinsky as Madonna is to .. Richard, perhaps your point is that IQ tests assume certain implicit background knowledge. I stated in my email that AIXI would equal any other intelligence starting with the same initial knowledge set So, your point is that IQ tests assume an initial knowledge set that is part and parcel of human culture. No, that was not my point at all. My point was much more subtle than that. You claim that AIXI would equal any other intelligence starting with the same initial knowledge set. I am focussing on the initial knowledge set. So let's compare me, as the other intelligence, with AIXI. What exactly is the same initial knowledge set that we are talking about here? Just the words I have heard and read in my lifetime? The words that I have heard, read AND spoken in my lifetime? The sum total of my sensory experiences, down at the neuron-firing level? The sum total of my sensory experiences AND my actions, down at the neuron firing level? All of the above, but also including the sum total of all my internal mental machinery, so as to relate the other fluxes of data in a coherent way? All of the above, but including all the cultural information that is stored out there in other minds, in my society? All of the above, but including simulations of all the related Where, exactly, does AIXI draw the line when it tries to emulate my performance on the test? (I picked that particular example of an IQ test question in order to highlight the way that some tests involve a huge amount of information that requires understanding other minds .. my goal being to force AIXI into having to go a long way to get its information). And if it does not draw a clear line around what same initial knowledge set means, but the process is open ended, what is to stop the AIXI theorems from implictly assuming that AIXI, if it needs to, can simulate my brain and the brains of all the other humans, in its attempt to do the optimisation? What I am asking (non-rhetorically) is a question about how far AIXI goes along that path. Do you know AIXI well enough to say? My understanding (poor though it is) is that it appears to allow itself the latitude to go that far if the optimization requires it. If it *does* allow itself that option, it would be parasitic on human intelligence, because it would effectively be simulating one in order to deconstruct it and use its knowledge to answer the questions. Can you say, definitively, that AIXI draws a clear line around the meaning of same initial knowledge set, and does not allow itself the option of implicitly simulating entire human minds as part of its infinite computation? Now, I do have a second line of argument in readiness, in case you can confirm that it really is strictly limited, but I don't think I need to use it. (In a nutshell, I would go on to say that if it does draw such a line, then I dispute that it really can be proved to perform as well as I do, because it redefines what I am trying to do in such a way as to weaken my performance, and then proves that it can perform better than *that*). 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] Apology to the list, and a more serious commentary on AIXI
Ben Goertzel wrote: I agree that, to compare humans versus AIXI on an IQ test in a fully fair way (that tests only intelligence rather than prior knowledge) would be hard, because there is no easy way to supply AIXI with the same initial knowledge state that the human has. Regarding whether AIXI, in order to solve an IQ test, would simulate the whole physical universe internally in order to simulate humans and thus figure out what a human would say for each question -- I really doubt it, actually. I am very close to certain that simulating a human is NOT the simplest possible way to create a software program scoring 100% on human-created IQ tests. So, the Occam prior embodied in AIXI would almost surely not cause it to take the strategy you suggest. -- Ben Alas, that was not quite the question at issue... In the proof of AIXI's ability to solve the IQ test, is AIXI *allowed* to go so far as to simulate most of the functionality of a human brain in order to acquire its ability? I am not asking you to make a judgment call on whether or not it would do so in practice, I am asking whether the structure of the proof allows that possibility to occur, should the contingencies of the world oblige it to do so. (I would also be tempted to question your judgment call, here, but I don't want to go that route :-)). If the proof allows even the possibility that AIXI will do this, then AIXI has an homunculus stashed away deep inside it (or at least, it has one on call and ready to go when needed). I only need the possibility that it will do this, and my conclusion holds. So: clear question. Does the proof implicitly allow it? 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] Apology to the list, and a more serious commentary on AIXI
Alas, that was not quite the question at issue... In the proof of AIXI's ability to solve the IQ test, is AIXI *allowed* to go so far as to simulate most of the functionality of a human brain in order to acquire its ability? I am not asking you to make a judgment call on whether or not it would do so in practice, I am asking whether the structure of the proof allows that possibility to occur, should the contingencies of the world oblige it to do so. (I would also be tempted to question your judgment call, here, but I don't want to go that route :-)). If the proof allows even the possibility that AIXI will do this, then AIXI has an homunculus stashed away deep inside it (or at least, it has one on call and ready to go when needed). I only need the possibility that it will do this, and my conclusion holds. So: clear question. Does the proof implicitly allow it? Yeah, if AIXI is given initial knowledge or experiential feedback that is in principle adequate for internal reconstruction of simulated humans ... then its learning algorithm may potentially construct simulated humans. However, it is not at all clear that, in order to do well on an IQ test, AIXI would need to be given enough background data or experiential feedback to **enable** accurate simulation of humans It's not right to way AIXI has a homunculus on call and ready to go when needed. Rather, it's right to say AIXI has the capability to synthesize an homunculus if it is given adequate data to infer the properties of one, and judges this the best way to approach the problem at hand. -- Ben G - 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] Apology to the list, and a more serious commentary on AIXI
Ben Goertzel wrote: Alas, that was not quite the question at issue... In the proof of AIXI's ability to solve the IQ test, is AIXI *allowed* to go so far as to simulate most of the functionality of a human brain in order to acquire its ability? I am not asking you to make a judgment call on whether or not it would do so in practice, I am asking whether the structure of the proof allows that possibility to occur, should the contingencies of the world oblige it to do so. (I would also be tempted to question your judgment call, here, but I don't want to go that route :-)). If the proof allows even the possibility that AIXI will do this, then AIXI has an homunculus stashed away deep inside it (or at least, it has one on call and ready to go when needed). I only need the possibility that it will do this, and my conclusion holds. So: clear question. Does the proof implicitly allow it? Yeah, if AIXI is given initial knowledge or experiential feedback that is in principle adequate for internal reconstruction of simulated humans ... then its learning algorithm may potentially construct simulated humans. However, it is not at all clear that, in order to do well on an IQ test, AIXI would need to be given enough background data or experiential feedback to **enable** accurate simulation of humans It's not right to way AIXI has a homunculus on call and ready to go when needed. Rather, it's right to say AIXI has the capability to synthesize an homunculus if it is given adequate data to infer the properties of one, and judges this the best way to approach the problem at hand. My overall argument is completely vindicated by what you say here. (My wording was sometimes ambiguous in that last email, I confess, but what I have been targeting is AIXI as proof, not AIXI as actual working system). I only care about where AIXI gets the power of its proof, so it does not matter to me whether a practical implementation [sic] of AIXI would actually need to build a cognitive system. It is not important whether it would do so in practice, because if the proof says that AIXI is allowed to build a complete cognitive system in the course of solving the IQ test problem, then what is the meaning of AIXI would equal any other intelligence starting with the same initial knowledge set well, yeah, of course it would, if it was allowed to build something as sophisticated as that other intelligence! It is like me saying I can prove that I can make a jet airliner with my bare hands . and then when you delve into the proof you find that my definition of make includes the act of putting in a phone call to Boeing and asking them to deliver one. Such a proof is completely valueless. AIXI is valueless. QED. 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] Apology to the list, and a more serious commentary on AIXI
AIXI is valueless. Well, I agree that AIXI provides zero useful practical guidance to those of us working on practical AGI systems. However, as I clarified in a prior longer post, saying that mathematics is valueless is always a risky proposition. Statements of this nature have been proved wrong plenty of times in the past, in spite of their apparent sensibleness at the time of utterance... But I think we have all made our views on this topic rather clear, at this point ;-) Time to agree to disagree and move on... -- 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] Apology to the list, and a more serious commentary on AIXI
On 3/8/07, Richard Loosemore [EMAIL PROTECTED] wrote: Let me put it this way: would AIXI, in building an implementation of this function, have to make use of a universe (or universe simulation) that *implicitly* included intelligences that were capable of creating the IQ tests? So, if there were a question like this in the IQ tests: Anna Nicole is to Monica Lewinsky as Madonna is to .. Would AIXI have to build a solution by implicitly deconstructing (if you see what I mean) the entire real universe, including its real human societies and real (intelligent) human beings and real social relationships? If AIXI does a post-hoc deconstruction of some real intelligent systems as part of building its own intelligent function, it is parasitic on that intelligence. You can confirm that it is not parasitic in that way? If I understand you correctly, you ask two different questions here. (Context: I'm assuming IQ test means a folder of IQ tests you might actually buy from a real company today, not some hypothetical function of arbitrary complexity.) The first question is, consider the shortest program that would max out the test. Does it consist of: A) Start with the Big Bang, run 14 billion years, pick the Everett branch that evolved English-speaking humans, send a UFO to abduct the smartest human and present him with the test... (okay I'm being a little facetious but you get the idea), B) Some special-purpose hack that treats Anna Nicole etc as arbitrary symbols without any of the connotations they have to us, _and does not generalize to anything much other than IQ tests_. Obviously it's unprovable, but I'm confident the answer is B based on experience: the shortest program for any _particular_ task is almost always a special-purpose hack that doesn't generalize. And in case B, everyone would agree there is no great intelligence involved. You then seem to be saying that even in case A, the intelligence would reside in the genius evolved in the simulated universe, and the apparent intelligence of AIXI would be parasitical on that, i.e. AIXI itself wouldn't really be intelligent. As I said recently, I agree with that position, but it's also one of philosophy, of how one chooses to define the word intelligence, not something amenable to proof or disproof; to call AIXI intelligent in that scenario would effectively be a form of pantheism. - 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] Apology to the list, and a more serious commentary on AIXI
On 3/8/07, Peter Voss [EMAIL PROTECTED] wrote: It's about time that some else said that the AIXI emperor has no clothes. Infinite computing power arguments prove **nothing**. That depends on exactly what you mean by prove nothing. For example, you can use the AIXI model to prove that no real computable AGI (I'm talking finite computing power now) is able to solve certain kinds of learning tasks. I consider that a real result, but I guess you do not. 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] Apology to the list, and a more serious commentary on AIXI
On 3/8/07, Peter Voss [EMAIL PROTECTED] wrote: AIXI certainly doesn't prove that AGI is possible. I agree. The human brain is what makes me think that it's possible. 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