jsmith@king.mcs.drexel.edu (Justin Smith) (07/03/90)
\hsize=6in
\hoffset=.4in
\centerline {\it The Possible Limitations of Artificial
Intelligence.}
\medskip
\centerline{by}
\medskip
\centerline{Justin R. Smith}
Roger Penrose has suggested that the human brain
has properties that may enable it to carry out actions
that are not reproducible by any computer. This
argument is used to imply that attempts to simulate the
reasoning cabability of the human mind mechanically are
essentially {\it futile}. His argument makes use of human
consciousness. I contend that one can come to the same
conclusion without appealing to human consciousness.
The basic idea is:
\item {1.} that the human brain is a {\it physical
object}.
\item {2.} Physical objects have the potential for
performing activities that are not reproducible by a
computer.
We will couch this in the terms of Computability Theory.
Consider two classes of functions defined (for the sake of
simplicity) with domain and range in the integers:
{\it recursive} functions; and {\it physical}
functions.
{\it Recursive functions} are essentially functions that
can be computed by executing a {\it computer program} of
some kind.
{\it Physical functions} are functions whose evaluation is
the result of observing some {\it physical process}.
An example of this is the number of ticks on a geiger
counter per minute as a function of time.
\proclaim{Claim}. The set of physical functions includes
the set of recursive functions.
This follows from the existence of physical
devices that are excellent {\it simulators} of Turing
machines --- I am using one to type this news item.
On the other hand, it is quite likely that the set of
physical functions is {\it strictly larger} than the set
of recursive functions. In fact,
quantum-mechanical phenomena suggest {\it precisely
this}. Quantum mechanics contains many manifestations
of ``random'' phenomena --- basically contending that
certain physical phenomena can only be analyzed {\it
statistically}. One can interpret ``random'' as meaning
``not computable'' rather than ``entirely devoid of
meaning''.
The human brain, being physical, has a {\it natural
tendancy} to make use of {\it physical functions} rather
than recursive functions in its computations.
Over the course of evolution (and we have to include the
evolution of the reptilian and mammalian as well as the
human brain) any physical functions that gave rise to
useful information {\it were utilized}. A rat fleeing
from a predator didn't ask whether the decision to flee
was the result of a recursive function evaluation.
The human brain wasn't designed by engineers who have an
interest in {\it filtering out} physical phenomena that
cause it to {\it depart} from strict turing-machine
computations (i.e., the effects of random thermal noise).
This is the only reasonable policy to follow in designing
computers
--- no engineer (nor anyone else, for that matter)
knows enough physics to ``program'' physical phenomena
{\it fully}. By this I mean: if ``random'' atomic
transitions turn out to really {\it mean something} we
don't know {\it what} they mean, or how to {\it exploit}
this ``information'' to solve problems.
The brain, on the other hand, has tens of millions of
years of ``experience'' at attempting to survive by any
means at its disposal, and it appears {\it likely} that
it makes use of physical computations that are {\it not}
Turing-computable.
I feel, that if we must regard the brain as a ``computer
program'', we have to concede that it uses {\it many
oracles} (in the sense of computability
theory)
\footnote*{Computability theory is concerned (among
other things) with: a. the question of what {\it is}
Turing-computable and, b. if one is {\it magically given}
information that might {\it not} be Turing-computable
(such a source of information is called an {\it oracle})
what {\it other} conclusions can one {\it derive} from
this source via Turing-machine-type computations. (I.e.,
given two recursively unsolvable problems, can a solution
to {\it one} be {\it recursively transformed} into a
solution of the other).}. Even the overall high-level {\it
control mechanism} of the brain may be a physical program
that isn't Turing computable. \enddaryl@oravax.UUCP (Steven Daryl McCullough) (07/03/90)
In article <1990Jul2.182411.4441@king.mcs.drexel.edu>, jsmith@king.mcs.drexel.edu (Justin Smith) writes: > The basic idea is: > \item {1.} that the human brain is a {\it physical > object}. > \item {2.} Physical objects have the potential for > performing activities that are not reproducible by a > computer. > [...stuff deleted...] Justin, your argument, though correct in a certain sense, doesn't address the issue of artificial intelligence at all, in my humble opinion. It is certainly true that because of the inherent randomness of quantum mechanics it is possible to create a physical process which does something uncomputable. For instance, a random number generator that uses radioactive decay will produce a sequence of numbers that almost certainly would not be produced by any given Turing machine program. However, why do you (or Penrose, for that matter) think that such randomness has anything to do with consciousness? It doesn't seem to contribute anything usefully noncomputable; for example, human beings cannot solve the halting problem any more than Turing machines can. Daryl McCullough
ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) (07/04/90)
In article <1990Jul2.182411.4441@king.mcs.drexel.edu> jsmith@king.mcs.drexel.edu (Justin Smith) writes: > >Intelligence.} >\medskip >Roger Penrose has suggested that the human brain >has properties that may enable it to carry out actions >that are not reproducible by any computer. This >argument is used to imply that attempts to simulate the >reasoning cabability of the human mind mechanically are >essentially {\it futile}. His argument makes use of human >consciousness. I contend that one can come to the same >conclusion without appealing to human consciousness. > >The basic idea is: >\item {1.} that the human brain is a {\it physical >object}. >\item {2.} Physical objects have the potential for >performing activities that are not reproducible by a >computer. >{\it Recursive functions} are essentially functions that >can be computed by executing a {\it computer program} of >some kind. > >{\it Physical functions} are functions whose evaluation is >the result of observing some {\it physical process}. >An example of this is the number of ticks on a geiger >counter per minute as a function of time. Good enough...but in the real world, one needs a physical device to run a computer programs. Thus all computer programs, when executed on a real computer, become physical functions. This means that claims based on a computer's inability to perform physical functions are flawed. >In fact, >quantum-mechanical phenomena suggest {\it precisely >this}. Quantum mechanics contains many manifestations >of ``random'' phenomena --- basically contending that >certain physical phenomena can only be analyzed {\it >statistically}. One can interpret ``random'' as meaning >``not computable'' rather than ``entirely devoid of >meaning''. There are more than one interpretations of the meanings of quantum-mechanical functions. Some people want them to be the fingers of God. Other people see them as the most "chaotic" functions, being utterly unpredictable. Whether QM functions can be computed by a Turing Machine has not to my knowledge been explicitly proven, and probably will never be. One thing for sure is that QM functions, being physical functions, can be determined by physical devices. Computers and brains are both physical devices. >The human brain, being physical, has a {\it natural >tendancy} to make use of {\it physical functions} rather >than recursive functions in its computations. >Over the course of evolution (and we have to include the >evolution of the reptilian and mammalian as well as the >human brain) any physical functions that gave rise to >useful information {\it were utilized}. Let us assume we have a genetic algorithm program running on a physical computer. It too will "evolve" utilizing whatever computational resources the programmer gives it. This may include an external Gieger Counter hooked up to the machine if you insist on having QM functions neccessary for intelligence. >The human brain wasn't designed by engineers who have an >interest in {\it filtering out} physical phenomena that >cause it to {\it depart} from strict turing-machine >computations (i.e., the effects of random thermal noise). >This is the only reasonable policy to follow in designing >computers > --- no engineer (nor anyone else, for that matter) >knows enough physics to ``program'' physical phenomena >{\it fully}. By this I mean: if ``random'' atomic >transitions turn out to really {\it mean something} we >don't know {\it what} they mean, or how to {\it exploit} >this ``information'' to solve problems. How do you reconcile the above statement with the below statement? >The brain, on the other hand, has tens of millions of >years of ``experience'' at attempting to survive by any >means at its disposal, and it appears {\it likely} that >it makes use of physical computations that are {\it not} >Turing-computable. If there is information yielded by QM functions, it can be determined by learning functions such as genetic algorithms, symbolic machine learning methods, or neural network functions such as backpropagation....I can't see how one can argue there is "hidden information" in QM functions which can only be interpreted by human evolution and not by any other learning system. Further, there are computer programs which use stochastic properties to make decisions (i.e. Simulated Annealing). I see absolutely NO PROOF that QM functions provide any useful information to an intelligent system which can be utilized. I see plenty of evidence that QM functions can be used like any other "random" function to provide probability spectra for stochastic decisions. I don't see proof why QM functions provide any advantage over chaotic functions with similar probability spectra. Sorry to be antagonistic, but I don't see why people can't accept the fact that brain is a physical computing device, as a digital computer is a physical computing device. The difference is that the brain relies on parallel non-linear computational methods on a scale we are 5 or more orders of magnitude away from, and has complex learning and organizational of sorts that connectionists are not even dreaming of yet. And a final note...just because a computer is digital does not mean it cannot perform parallel analogue equations. It might be limited by the "quanta" of it's least significant bit, but so too are chemical reactions in brain limited by the "quanta" of chemical molecules, and electric phenomena in brain limited by "quanta" of a single electron. Real valued functions in the real world have the same quantification problems that real values have on computers (though there are alot more significant bits in the real world :-). On a side note, I just completed training a neural net to recognize valid targets from IR focal plane arrays. All I can say is that the network learned alot more about categorizing valid targets from invalid targets than I did (I didn't even look at most of the data). -Thomas Edwards
als@bohra.cpg.oz (Anthony Shipman) (07/04/90)
In article <1990Jul2.182411.4441@king.mcs.drexel.edu>, jsmith@king.mcs.drexel.edu (Justin Smith) writes: > The human brain, being physical, has a {\it natural > tendancy} to make use of {\it physical functions} rather > than recursive functions in its computations. > Over the course of evolution (and we have to include the > evolution of the reptilian and mammalian as well as the > human brain) any physical functions that gave rise to > useful information {\it were utilized}. A rat fleeing > from a predator didn't ask whether the decision to flee > was the result of a recursive function evaluation. > > > The brain, on the other hand, has tens of millions of > years of ``experience'' at attempting to survive by any > means at its disposal, and it appears {\it likely} that > it makes use of physical computations that are {\it not} > Turing-computable. Implicit in all of these types of arguments is the assumption that whatever the brain uses to achieve intelligence is the one and only way it can be done. I consider this to be an unjustified assumption. Counterargument: No machine does nor can flap its wings well enough to fly but many fly nonetheless. And better than birds do. Since nobody understands: intelligence, how-the-brain-works, knowledge, meaning, understanding etc. all arguments about whether AI is possible or not are just mind games IMHO. Maybe in 50 or 100 years we may know enough about the subject to carry out a more knowledgeable discussion. In the worst possible case searching for AI may be like searching for the philospher's stone. This was an unachievable goal but along the way a great deal of useful knowledge was obtained. Similarly I believe the search for AI will be a fruitful task even if the end goal turns out to be unachievable. To even seek to abort this task at this early stage is incredibly myopic. -- Anthony Shipman ACSnet: als@bohra.cpg.oz.au Computer Power Group 9th Flr, 616 St. Kilda Rd., St. Kilda, Melbourne, Australia D
dg1v+@andrew.cmu.edu (David Greene) (07/05/90)
Excerpts from netnews.comp.ai: 4-Jul-90 Re: Artificial vs. ''real''.. Anthony Shipman@bohra.cp (2032) > Counterargument: No machine does nor can flap its wings well enough to > fly but > many fly nonetheless. And better than birds do. I don't disagree with your post and I suspect your "counterargument" was meant lightly, however, it highlights an important point: The machines that man built to fly serve a different purpose than those in nature -- I don't see man's as inherently "better". Just like mainframes that can do a higher volumes of computation much faster, so to can jet planes fly higher volumes faster... the problem for AI is building a machine that is small and fast enough to flit from tree to tree dodging whatever lies in between. -David -------------------------------------------------------------------- David Perry Greene || ARPA: dg1v@andrew.cmu.edu GSIA /Robotics || dpg@isl1.ri.cmu.edu Carnegie Mellon Univ. || BITNET: dg1v%andrew@vb.cc.cmu.edu Pittsburgh, PA 15213 || UUCP: !harvard!andrew.cmu.edu!dg1v -------------------------------------------------------------------- "You're welcome to use my opinions, just don't get them all wrinkled."
dsa@dlogics.COM (David Angulo) (07/07/90)
In article <5734@jhunix.HCF.JHU.EDU>, ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) writes: > > Whether QM functions can be computed by a Turing Machine has not > to my knowledge been explicitly proven, and probably will never > be. QM "functions" cannot be "computed" by any means. All you can "compute" are amplitudes and probability densities. > > One thing for sure is that QM functions, being physical functions, > can be determined by physical devices. No they cannot. Please stop saying this. It is incorrect as has been pointed out here many times. > This may include > an external Gieger Counter hooked up to the machine if you > insist on having QM functions neccessary for intelligence Well, I don't insist that QM is necessary for intelligence (outside of how it is important for the world); however, to say something like this you need to understand QM better. "Hooking up a Geiger Counter" will probably change the result of your experiment. > -- David S. Angulo (312) 266-3134 Datalogics Internet: dsa@dlogics.com 441 W. Huron UUCP: ..!uunet!dlogics!dsa Chicago, Il. 60610 FAX: (312) 266-4473
tim@cstr.ed.ac.uk (Tim Bradshaw) (07/07/90)
>>>>> On 6 Jul 90 21:17:48 GMT, dsa@dlogics.COM (David Angulo) said: > In article <5734@jhunix.HCF.JHU.EDU>, ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) writes: >> >> One thing for sure is that QM functions, being physical functions, >> can be determined by physical devices. > No they cannot. Please stop saying this. It is incorrect as has been pointed > out here many times. Is this true? I have a paper by Deutch where he proves something he calls the `physical Church-Turing principle': Every finitely realizable physical system can be perfectly simulated by a universal model computing machine operating by finite means. For QM & something he calls a `universal quantum computer'. He also demonstrates that such a machine is in theory possible to construct & that it has many interesting properties. Note that this principle is *not* true for classical mechanics and the conventional universal Turing machine. I think that it is (one of) Penrose's ideas that the brain may be a quantum computer. I also think that this is unlikely, actually, since it is too big & too hot. All it seems likely to be able to rely on is some sort of random oracle & one can easily add this to a normal Turing machine. Apologies if this goes over old ground. --tim Tim Bradshaw. Internet: tim%ed.cstr@nsfnet-relay.ac.uk UUCP: ...!uunet!mcvax!ukc!cstr!tim JANET: tim@uk.ac.ed.cstr "Quis custodiet ipsos custodes?"
ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) (07/08/90)
In article <598@dlogics.COM> dsa@dlogics.COM (David Angulo) writes: >In article <5734@jhunix.HCF.JHU.EDU>, ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) writes: >> One thing for sure is that QM functions, being physical functions, >> can be determined by physical devices. >No they cannot. Please stop saying this. It is incorrect as has been pointed >out here many times. I think we have a misunderstanding here. When I say "determined", I mean measured. For example, an electron passing through a diffraction grating has a probability field of where it will end up hitting a target. There is no evidence that any a priori method can discover where exactly it will hit. The electron collision location can be located by physical devices after the collision. _That's_ what I meant by "determined" and "computed", not pre_determination or Turing Computable (My words were a little misleading...sorry). There are some who believe that this apparent randomness is the "Philosopher's Stone" of intelligence...that some mysterious "force" can make the electron go (in the case stated above) to the proper location on the target which might enable a system utilizing electron collision detectors on the target to make an "intelligent" decision. (Of course, those who hold to this tenet feel that this happens in a real neural system, not our diffraction grating). >> This may include >> an external Gieger Counter hooked up to the machine if you >> insist on having QM functions neccessary for intelligence >Well, I don't insist that QM is necessary for intelligence (outside of >how it is important for the world); however, to say something like this >you need to understand QM better. "Hooking up a Geiger Counter" will >probably change the result of your experiment. Exactly. That is the fallicy behind "QM-induced intelligence." Now, I see no reason why QM-probability fields cannot be used for stochastic computations, but there is no significant benefit in using it over other well-behaved random systems. -Thomas Edwards
dsa@dlogics.COM (David Angulo) (07/10/90)
In article <5767@jhunix.HCF.JHU.EDU>, ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) writes: > In article <598@dlogics.COM> dsa@dlogics.COM (David Angulo) writes: > >In article <5734@jhunix.HCF.JHU.EDU>, ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) writes: > > >> One thing for sure is that QM functions, being physical functions, > >> can be determined by physical devices. > > >No they cannot. Please stop saying this. It is incorrect as has been pointed > >out here many times. > > I think we have a misunderstanding here. When I say "determined", I > mean measured. Sorry, you can't measure them either. That is, you can measure a particle's physical position (with, say your Geiger counter) but then you don't know[ where it is going or where it has been. Or you can measure its momentum but then you don't know where it is. > For example, an electron passing through a diffraction > grating has a probability field of where it will end up hitting > a target. There is no evidence that any a priori method can discover > where exactly it will hit. The electron collision location can be > located by physical devices after the collision. What if you send it through two slits? Then it actually was in two points "at once!" And it will interfere with itself. You cannot measure where it is at all. If you do, it will change your experiment. It will no longer interfere with itself. It no longer went through both slits. QM is difficult, I'll grant you but just try to think of these "particles" as things that do not behave as what we intuitively understand as particles behaving (always, anyway). Also, there is no "probability field." There is a wave equation with which is associated an amplitude. You can use this to compute a probability density but this is not a field. -- David S. Angulo (312) 266-3134 Datalogics Internet: dsa@dlogics.com 441 W. Huron UUCP: ..!uunet!dlogics!dsa Chicago, Il. 60610 FAX: (312) 266-4473
ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) (07/11/90)
In article <599@dlogics.COM> dsa@dlogics.COM (David Angulo) writes: >In article <5767@jhunix.HCF.JHU.EDU>, ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) writes: >> I think we have a misunderstanding here. When I say "determined", I >> mean measured. > >Sorry, you can't measure them either. That is, you can measure a particle's >physical position (with, say your Geiger counter) but then you don't know[ >where it is going or where it has been. Or you can measure its momentum >but then you don't know where it is. OK, let's replace "measuring a QM function" (which I accept is impossible) with measuring an aspect of an object (such as velocity or position). This does not preclude the fact that if I measure one aspect of an object, I neccessarily 'change the experiment' and change other aspects of the object. Anyway, for QM phenomena to have some relevance to real neural system computation, there must be "measurement" of one aspect of the QM phenomena. I still state that I see no reason why people feel that there are some kinds of QM phenomena being measured in the brain which gives real neural systems intelligence which cannot be replicated by artificial systems. We also see above that Dave points out that this measurement of one aspect of QM phenomena neccessarily effects other aspects of the QM phenomena. I don't see how this validates or invalidates the claim of QM based intelligence (maybe Dave can point that out). -Thomas
dsa@dlogics.COM (David Angulo) (07/11/90)
In article <TIM.90Jul7140549@watt.cstr.ed.ac.uk>, tim@cstr.ed.ac.uk (Tim Bradshaw) writes: > I have a paper by Deutch where he proves something he > calls the `physical Church-Turing principle': > > Every finitely realizable physical system can be perfectly > simulated by a universal model computing machine operating by > finite means. > Well, I only majored in Physics in college. I do not even pretend to be an "expert." Also, I haven't read this reference and perhaps you're misquoting it; however, I don't think that this is possible as current physical theories do not explain everything known about the universe so if this guy can perfectly model the universe I would think that the physics community would look on him like a god. Just to show the complexities of QM theory (and this is a WELL UNDERSTOOD phenomenon not one of those as yet unexplainable ones), if you put a particle (let's call it an electron just so we can picture it) into a box and made the walls of the box such that it would take an infinate amount of energy to get the particle past that wall, with our a priori notions of matter, we would expect that the particle is doomed to stay in the box for all time. However, if we work out the wave equation for that particle and use that to calculate the probability densities over the spatial coordinates, we find that the particle has a finite probability of being outside of the box. This is not just theoretical. This is the basis of the principle of tunnelling on which all semiconductor technology is based. Maybe you can clarify what this "Deutch" was talking about and when he said it? -- David S. Angulo (312) 266-3134 Datalogics Internet: dsa@dlogics.com 441 W. Huron UUCP: ..!uunet!dlogics!dsa Chicago, Il. 60610 FAX: (312) 266-4473
dave@hpgnd.HP.COM (Dave PENKLER) (07/11/90)
In article <1990Jul2.182411.4441@king.mcs.drexel.edu>, jsmith@king.mcs.drexel.edu (Justin Smith) writes: > Roger Penrose has suggested that the human brain > has properties that may enable it to carry out actions > that are not reproducible by any computer. This > argument is used to imply that attempts to simulate the > reasoning capability of the human mind mechanically are > essentially {\it futile}. His argument makes use of human > consciousness. I contend that one can come to the same > conclusion without appealing to human consciousness. > [ stuff omitted ] > > On the other hand, it is quite likely that the set of > physical functions is {\it strictly larger} than the set > of recursive functions. In fact, > quantum-mechanical phenomena suggest {\it precisely > this}. Quantum mechanics contains many manifestations > of ``random'' phenomena --- basically contending that > certain physical phenomena can only be analyzed {\it > statistically}. One can interpret ``random'' as meaning > ``not computable'' rather than ``entirely devoid of > meaning''. [ more stuff omitted ] > I feel, that if we must regard the brain as a ``computer > program'', we have to concede that it uses {\it many > oracles} (in the sense of computability > theory) A book that I read called 'The Spirit of Matter' gives an interesting if not very conjectural theory of the 'oracles' mentioned by Justin (vide supra). The author (I forget the name) is a theoretical physicist who developed a complex formulation of Einstein's Relativity Theory. The gist of the book, very roughly, is as follows Electrons are indeed very old, the large majority anyway, since spontaneous electron (positron) creation by collision of very high energy photons is rare. The density of an electron is such that it can be considered to be a tiny black-hole in its own right. When a photon impinges on an electron it is subjected to the relativistic effects of the little black-hole and as such never actually gets there. Energy conservation is kept by in a change of orbital. When the electron drops back it releases a photon. The direction of propagation of this released photon is *not* predictable. The author of the book claims that the direction of the released photon is determined by the aggregate configuration of photons still 'arriving' at the electron since its creation. In this way the electrons 'communicate'. Because the electron is a closed system it never 'loses' information, i.e. inside it entropy only decreases = neg-entropy increases. (tenuous identification of neg-entropy i.e. order and information). Given that the bulk of electrons in the matter that constitutes our world are a old as the universe, that there is a kind of _action-at-a-distance_ he concludes that matter itself is the base of intelligence, a system capable of evolving into galaxies, planets, life and indeed into organisms capable forming theories on these things. (BTW this theory is not only limited to the leptons but can also plausibly be extended to the hadrons) Now looking at the problem of A.I. in this light, the sheer volume of information and the number of interactions occurring that engender 'Real intelligence' makes the job of simulating it about as big as re-creating the universe itself. (How many states does this beastie have ? How many state transitions occur per second ? The number of states increases with each photon emitted and the sun is just an average star.) I don't think we want to simulate nature (building birds that can flit from tree to tree) so that we can take our place on Mount Olympus, but to learn from nature so that we can facilitate our lives and give ourselves the time and means to do the things we want. For me artificial intelligence involves the adaptation, implementation, application and exploitation of processes, assimilated from nature, deemed by us the engineers as intelligent. The adoption of suitable epistemological frameworks is an important consideration in the field of A.I. but we should leave the BIG PICTURE for the philosophers to ponder about. -Dave PENKLER "Objects are nothing but debilitated functional-values"
jsmith@king.mcs.drexel.edu (Justin Smith) (07/11/90)
In article <TIM.90Jul7140549@watt.cstr.ed.ac.uk>, tim@cstr.ed.ac.uk (Tim Bradsha w) writes: > I have a paper by Deutch where he proves something he > calls the `physical Church-Turing principle': > > Every finitely realizable physical system can be perfectly > simulated by a universal model computing machine operating by > finite means. > The key element here is ''finitely realizable''. There is no reason to assume that ''finitely realizable'' physical systems exist!
dsa@dlogics.COM (David Angulo) (07/12/90)
In article <5781@jhunix.HCF.JHU.EDU>, ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) writes: > OK, let's replace "measuring a QM function" (which I accept is impossible) > with measuring an aspect of an object (such as velocity or position). Which is really pointless. Why would you want to measure its momentum if you have no idea where in the universe it is (that is, the particle would exist as a pure wave function, sort of). > > Anyway, for QM phenomena to have some relevance to real > neural system computation, there must be "measurement" of one > aspect of the QM phenomena. > Those who argue that neural systems are based on QM phenoment do not argue this way, I believe. I don't want to talk to them. > I still state that I see no reason why people feel that > there are some kinds of QM phenomena being measured in the brain > which gives real neural systems intelligence which cannot be > replicated by artificial systems. > Well, I think I do not like your use of the word "measure" because the brain is not doing any "measurements;" however, I don't see how the phenomenon can be made use by brain material any differently than artificial systems would. > We also see above that Dave points out that this measurement > of one aspect of QM phenomena neccessarily effects other > aspects of the QM phenomena. I don't see how this > validates or invalidates the claim of QM based intelligence (maybe Dave > can point that out). > I don't either. I basically agree with your premise. I just don't want us to be picked on by incorrect use of QM. I believe (as you do) that Penrose, et. al. are using QM incorrectly. -- David S. Angulo (312) 266-3134 Datalogics Internet: dsa@dlogics.com 441 W. Huron UUCP: ..!uunet!dlogics!dsa Chicago, Il. 60610 FAX: (312) 266-4473
tim@cstr.ed.ac.uk (Tim Bradshaw) (07/12/90)
>>>>> On 10 Jul 90 20:05:40 GMT, dsa@dlogics.COM (David Angulo) said: > In article <TIM.90Jul7140549@watt.cstr.ed.ac.uk>, tim@cstr.ed.ac.uk (Tim Bradshaw) writes: >> I have a paper by Deutch where he proves something he >> calls the `physical Church-Turing principle': >> >> Every finitely realizable physical system can be perfectly >> simulated by a universal model computing machine operating by >> finite means. >> > Well, I only majored in Physics in college. I do not even pretend to be > an "expert." Also, I haven't read this reference and perhaps you're > misquoting it; however, I don't think that this is possible as current > physical theories do not explain everything known about the universe so > if this guy can perfectly model the universe I would think that the > physics community would look on him like a god. No, this isn't what he's saying, and yes, I should have made it clearer. He's interested in the functions that can be computed by a physical system or a computing machine -- in the sense that, for instance, one can design a classical physical system that will calculate sin(x) given x, but you cannot write a program for a Turing machine which will do this. Loosely he says that a computing machine `perfectly simulates' a physical system if a mapping can be set up such that they calculate the same functions for a given program on the machine. His claim is then that for quantum mechanical systems that obey certain fairly general and plausible conditions then there exists some program under which the universal quantum computer will perfectly simulate such a system. This is *not* the same as giving the program of course -- all he's saying is that a program exists, which is a much lesser claim. This is in Proc Roy Soc London sometime early eighties (paper's at home). --tim Tim Bradshaw. Internet: tim%ed.cstr@nsfnet-relay.ac.uk UUCP: ...!uunet!mcvax!ukc!cstr!tim JANET: tim@uk.ac.ed.cstr "Quis custodiet ipsos custodes?"
dsa@dlogics.COM (David Angulo) (07/12/90)
In article <601@dlogics.COM>, dsa@dlogics.COM (David Angulo) writes: > Those who argue that neural systems are based on QM phenoment do not > argue this way, I believe. I don't want to talk to them. > Sorry, that was a typo. It should have read, "I don't want to talk for them." -- David S. Angulo (312) 266-3134 Datalogics Internet: dsa@dlogics.com 441 W. Huron UUCP: ..!uunet!dlogics!dsa Chicago, Il. 60610 FAX: (312) 266-4473
dsa@dlogics.COM (David Angulo) (07/14/90)
In article <TIM.90Jul11214014@watt.cstr.ed.ac.uk>, tim@cstr.ed.ac.uk (Tim Bradshaw) writes: > > He's interested in the functions that can be computed by a physical > system or a computing machine -- in the sense that, for instance, one > can design a classical physical system that will calculate sin(x) > given x, but you cannot write a program for a Turing machine which > will do this. Loosely he says that a computing machine `perfectly > simulates' a physical system if a mapping can be set up such that they > calculate the same functions for a given program on the machine. His > claim is then that for quantum mechanical systems that obey certain > fairly general and plausible conditions then there exists some program > under which the universal quantum computer will perfectly simulate > such a system. OK, I think I'm getting a clearer picture on what he's saying. It sounds as if his basic misunderstanding comes in simulating wave equations. You can simulate a sin() function because it really exists in the physical universe. Whether or not Schroedinger wave equations exist is still an area of contention. It is the RESULT of these wave equations that can be used. When you do this, you end up with a probability density. This CANNOT be "simulated" because it doesn't pertain to reality. It does not inform us that a particle exists at a certain point at a certain time. It only tells us what the probability of finding a particle at a certain point at a certain time is. -- David S. Angulo (312) 266-3134 Datalogics Internet: dsa@dlogics.com 441 W. Huron UUCP: ..!uunet!dlogics!dsa Chicago, Il. 60610 FAX: (312) 266-4473
lynch@CERC.UTEXAS.EDU (Tom Lynch) (07/17/90)
To Net land,
I have been trying to post this since early July, my apologizes
for repeated material ... If someone, by some strange hapenstance,
sees this on comp.ai, please send me some mail.
Penrose, and Searle have claimed that 'understanding' is not
Turing computable. Anyway the thought occurs to me that a
random process may be important in machine learning .. and AI in general.
An interesting phenomena in caches is that random replacement
has similar performance to least recently used. This is well
known. A friend mentioned a similar statistic in
knowledge bases -- random forgetting has similar performance to
weighted least recently used forgetting!
More specifically, in the microprocessor instruction cache
problem there is a fast on chip instruction memory that fails to
contain a required instruction. So, the required instruction
must be fetched from the slow external memory. When the required
instruction arrives at microprocessor it must replace something
already in the cache. A good replacement policy is to write the
new required instruction over the least recently used instruction
in the cache. In other words, forget the instruction that hasn't
been used for the longest amount of time. Randomly picking an
instruction to replace works just as well -- as was found and
implemented on some real microprocessors. It is not the case
that replacement policies are just bad -- caches can achieve very
high hit rates with either random or least recently used
replacement!
Apparently a similar phenomena exists in machine learning when
picking which rule in a knowledge base to forget, in order to
decrease search time. Apparently some knowledge bases will work
more efficiently if they don't "know too much". This sounds
intuitive - at some point search time must overwhelm the benefit
of knowing more. When something new and important must replace
something already known I was surprised to hear that a popular
algorithm was the same as for instruction caches - throw out the
least recently used knowledge. It was also surprising to learn
that RANDOM forgetting was almost as good as least recently used!
It is very interesting that a random decision can be
better than many 'thoughtful' decisions - and as good as at least
one thoughtful decision! I think this counter intuitive phenomena
may be important. Perhaps the random forgetting and the
impreciseness of human cognition and learning may not just be
important - but be REQUIRED.
Anyway wouldn't it be interesting to propose a turing machine
which has a randomly corrupted tape? There would be some finite
probability that the information on the tape would change.
Such a machine would not necessarily become random in its control
sequences, since the control unit could recover from a small
number of errors (although a small probability of failure would
exist). I bet the machine would actually become 'smarter'.
Would 'understanding' be computable by Penrose's argument with a
machine like this? I am calling the machine a Political Machine
since it is corrupted :-). Do you know how to go about finding
the language accepted by a Political Machine?
-tom lynch@cerc.utexas.edu
tim@cstr.ed.ac.uk (Tim Bradshaw) (07/28/90)
>>>>> On 13 Jul 90 18:34:42 GMT, dsa@dlogics.COM (David Angulo) said: > OK, I think I'm getting a clearer picture on what he's saying. It sounds > as if his basic misunderstanding comes in simulating wave equations. You > can simulate a sin() function because it really exists in the physical > universe. Whether or not Schroedinger wave equations exist is still > an area of contention. It is the RESULT of these wave equations that can > be used. When you do this, you end up with a probability density. This > CANNOT be "simulated" because it doesn't pertain to reality. It does not > inform us that a particle exists at a certain point at a certain time. It > only tells us what the probability of finding a particle at a certain point > at a certain time is. Sure, all he's saying is that you can perfectly simulate anything you can measure; and good quantum mechanists are only ever concerned with what you can measure: QM is really *about* what you can measure (or `observe' in the jargon...). Of course this raises the interesting question of what makes the observation, which has at least historically been a point of discussion among quantum theorists. The problem is that you must make an observation of a system to put it into a definite state (`collapse the wave-function/state-vector'), and this process is quite important in QM. Simple-mindedly you say that a `conscious observer' is what you need to make an observation in this sense. Of course, when you start getting interested in how to build such a conscious observer (which is one goal of AI in a sense) you have a problem because you're trying to construct a physical system which itself must obey QM, and just what is making the observation and where it's being made become obscure to say the least. Simple mindedly again, you can be lead to believe that an observer is not a physical system. The original solution to this was the Copenhagen interpretation, which really says `None of this matters, because you don't need it to do physics'. It seems that you would need it to understand what building a conscious observer is though. Well it's too long since I thought about this, but I think there are formulations of QM which get around this or have some hope of doing so. If anyone out there knows about this I'd like to hear from you, otherwise I'll have to look it up... --tim Tim Bradshaw. Internet: tim%ed.cstr@nsfnet-relay.ac.uk UUCP: ...!uunet!mcvax!ukc!cstr!tim JANET: tim@uk.ac.ed.cstr "Quis custodiet ipsos custodes?"