eitan@wisdom.BITNET.UUCP (10/30/87)
Had it ever come into you mind that simulating/emulating the human brain is
NP problem ? ( Why ? Think !!! ). Unless some smartass comes out with a proof
for NP=P yar can forget de whole damn thing ...
Eitan Shterenbaum
(*
As far as I know one can't solve NP problems even with a super-duper
hardware, so building such machine is pointless (Unless we are living on
such machine ...) !
*)
Eitanhonavar@SPEEDY.WISC.EDU (A Buggy AI Program) (10/31/87)
In article <8710280748.AA21340@jade.berkeley.edu> eitan@wisdom.BITNET (Eitan Shterenbaum) writes: > >Had it ever come into you mind that simulating/emulating the human brain is >NP problem ? ( Why ? Think !!! ). Unless some smartass comes out with a proof >for NP=P yar can forget de whole damn thing ... > > Eitan Shterenbaum >(* > As far as I know one can't solve NP problems even with a super-duper > hardware, so building such machine is pointless (Unless we are living on > such machine ...) ! >*) Discovering that a problem is NP-complete is usually just the beginning of the work on the problem. The knowledge that a problem is NP-complete provides valuable information on the lines of attack that have the greatest potential for success. We can concentrate on algorithms that are not guaranteed to run in polynomial time but do so most of the time or those that give approximate solutions in polynomial time. After all, the human brain does come up with approximate (reasonably good) solutions to a lot of the perceptual tasks although the solution may not always be the best possible. Knowing that a problem is NP-complete only tells us that the chances of finding a polynomial time solution are minimal (unless P=NP). -- VGH
eitan@WISDOM.BITNET (Eitan Shterenbaum) (11/05/87)
In article <> honavar@speedy.wisc.edu (A Buggy AI Program) writes: > >Discovering that a problem is NP-complete is usually just the >beginning of the work on the problem. The knowledge that a problem is >NP-complete provides valuable information on the lines of attack that >have the greatest potential for success. We can concentrate on algorithms >that are not guaranteed to run in polynomial time but do so most >of the time or those that give approximate solutions in polynomial time. >After all, the human brain does come up with approximate (reasonably good) >solutions to a lot of the perceptual tasks although the solution may not >always be the best possible. Knowing that a problem is NP-complete only >tells us that the chances of finding a polynomial time solution are minimal >(unless P=NP). > You are right and so am I, a) There're no polynomial algorithms, which are known to us, that can solve NP problems. b) There are approximate and probabilistic *partial* solutions for NP problems. As to the claim "the brain does it so why shouldn't the computer" - It seem to me that you forget that the brain is built slightly differently than a Von-Neuman machine ... It's a distributed enviorment lacking boolean algebra. I can hardly believe that even with all the partial solutions for all the complicated sets of NP problems that emulating a brain brings up, one might be able to present a working program. If you'd able to emulate mouse's brain you'd become a legend in your lifetime ! Anyway, no one can emulate a system which has no specifications. if the neuro-biologists would present them then you'd have something to start with. And last - Computers aren't meta-capable machines they have constraints, not every problem has an answer and not every answermakes sense, NP problems are the best example. Eitan Shterenbaum
honavar@SPEEDY.WISC.EDU (A Buggy AI Program) (11/09/87)
In article <4357@wisdom.BITNET> eitan%H@wiscvm.arpa (Eitan Shterenbaum) writes: > >As to the claim "the brain does it so why shouldn't the computer" - >It seem to me that you forget that the brain is built slightly differently >than a Von-Neuman machine ... It's a distributed enviorment lacking boolean >algebra. I can hardly believe that even with all the partial solutions for >all the complicated sets of NP problems that emulating a brain brings up, one >might be able to present a working program. If you'd able to emulate mouse's >brain you'd become a legend in your lifetime ! >Anyway, no one can emulate a system which has no specifications. >if the neuro-biologists would present them then you'd have something to start >with. I use the term "computer" in a sense somewhat broader than a Von-Neuman machine. We can, in principle, build machines that incorporate distributed representations, processing and control. It is not clear what you mean by a "distributed environment lacking boolean algebra." The use of fine-grained distributed representations naturally results in behavior indicative of processes using fuzzy or probabilistic logic. The goal is, not necessarily to emulate the brain in all its detail: We can study birds to understand the principles of aerodynamics that explain the phenomenon of flying and then go on to build an aeroplane that is very different from a bird but still obeys the same laws of physics. As for specifications, they can be provided in different forms and at different levels of detail; Part of the exercise is to discover such specifications - either by studying actual existing systems or by analyzing the functions needed at an abstract level to determine the basic building blocks and how they are to be put together. > >And last - Computers aren't meta-capable machines they have constraints, > not every problem has an answer and not every answermakes sense, > NP problems are the best example. > Are you implying that humans are "meta-capable" - whatever that means? VGH
eitan@WISDOM.BITNET.UUCP (11/15/87)
In article <> honavar@speedy.wisc.edu (A Buggy AI Program) writes: > >In article <4357@wisdom.BITNET> eitan%H@wiscvm.arpa (Eitan Shterenbaum) writes: >> >>Anyway, no one can emulate a system which has no specifications. >>if the neuro-biologists would present them then you'd have something to start >>with. > > I use the term "computer" in a sense somewhat broader than a > Von-Neuman machine. We can, in principle, build machines that ^^^^^^^^^^^^ ^^^^^^^^^^^^ > incorporate distributed representations, processing and control. > It is not clear what you mean by a "distributed environment lacking > boolean algebra." > The use of fine-grained distributed representations naturally results > in behavior indicative of processes using fuzzy or probabilistic logic. > The goal is, not necessarily to emulate the brain in all its detail: > We can study birds to understand the principles of aerodynamics that > explain the phenomenon of flying and then go on to build an aeroplane > that is very different from a bird but still obeys the same laws of > physics. As for specifications, they can be provided in different > forms and at different levels of detail; Part of the exercise is > to discover such specifications - either by studying actual existing > systems or by analyzing the functions needed at an abstract level to > determine the basic building blocks and how they are to be put > together. > a) You can't understand the laws under which a system works without understanding the structure of the system ( I believe that our intelligence is the result of our brain's structure ) b) The earodynamics example just prooves my point. Only after understanding *WHY* the birds are built in a certain form the researchers would've been able to understand the pronciples. The fact is that Leonardo de Vinci knew more about aerodynamics than the pioneers of flight is acknowlodged to the *research* he has done on birds. It seems to me that many AI scientists disregard 2 facts a- They have no definition of AI b- They disregard the fact that the best way to have more knowledge about a certain phenomennon is to observe and research it. It seems to me that 1) You have no definition for Intelligence. 2) You want to have the rules of Itelligence. 3) Thus you build systems inorder to simulate Intelligence. 4) Since you don't know you're looking for and since you have no basic rules to simulate the intelligence on, you invent your own local definition and rules for Intelligence. 5) Then youtry to mach your results with your expectations of what the results should be. Sometimes it works some time it doesn't. This method reminds me "random sort" I.E The computer has N numbers, It randomly prints them out one by one and then it tries to check whether they are ordered, if not - he does the above again. I hope that you've noticed that the probability that you'd be correct is quite slime ( actually 1/N! ... ) >> >>And last - Computers aren't meta-capable machines they have constraints, >> not every problem has an answer and not every answermakes sense, >> NP problems are the best example. >> > Are you implying that humans are "meta-capable" - whatever that means? > I'm trying to imply that human beings aren't Turing equivalent ... ( not even when compared to a non-determinitstic turing machine ) Correct me if I'm wrong but I do feel that the neuro-biologists chaps are in the right track and that the Computer scientists should combine efforts with them instead of messing around with AI. (I'm not saying that AI isn't usefull, it is, just that it's very little success in Inteligence and a grand success in Artificial artifacts ...) Eitan Shterenbaum Disclaimer - My ideas are mine and only mine ! @@@@@@@@@@@@@@@@@@@@@@@@@@@@