Newsgroups: comp.ai.alife,comp.ai.philosophy,comp.ai,alt.consciousness
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!ix.netcom.com!netcom.com!vlsi_lib
From: vlsi_lib@netcom.com (Gerard Malecki)
Subject: Re: Thought Question
Message-ID: <vlsi_libD2q5FC.FMG@netcom.com>
Organization: VLSI Libraries Incorporated
References: <3fmn70$cpa@pentagon.io.com> <vlsi_libD2oKoJ.GKw@netcom.com> <3forh6$7ub@pentagon.io.com>
Date: Fri, 20 Jan 1995 22:06:47 GMT
Lines: 186
Xref: glinda.oz.cs.cmu.edu comp.ai.alife:1888 comp.ai.philosophy:24842 comp.ai:26676

In article <3forh6$7ub@pentagon.io.com> larrys@pentagon.io.com (Larry Smith) writes:
>In article <vlsi_libD2oKoJ.GKw@netcom.com>,
>Gerard Malecki <vlsi_lib@netcom.com> wrote:
>>In article <3fmn70$cpa@pentagon.io.com> larrys@pentagon.io.com (Larry Smith) writes:
>>>
>>>Neurons have been simulated by a computer, which _is_ a Turing machine.
>>
>>This does not mean that neurons are themselves Turing machines. A neuron
>
>Err...yes, it does.  That's axiomatic, I'm afraid.  I don't care to go
>ito the proof here.
>
>>has a cretain degree of randomness to it, which may depend on quantum
>
>Turing machines are capable of sufficiently random behaviour to pass any
>test for it you care to name, even if the randomness _is_ algorithmic.
>The _source_ of randomness is utterly irrelevent for computation pur-
>poses, if the source can generate sufficiently random numbers.  Turning
>machines can be as random as needed, they can get asymptotically close
>to quantum variability.  You are looking for God under that curve, and
>He isn't there.
>
>>phenomena. A main characterisitic feature of Turing machines is repeatability
>>which is not the case with real-life neurons. In fact, a lot of processes
>
>I won't speak for _your_ neurons, but neurons are repeatable within
>the design limits of their construction.  The simulation will not
>wear out or overload, as a real neuron can, but so long as that does
>not happen a neuron _is_ repeatable.  Bear in mind that neural nets
>exhibit a lot of what is now called "chaos", the patterns in networks
>are not directly repeatable, but are predictable and simulatable
>in principle, using chaos theory.
>
>>such as thermal noise in a resistor are of true random nature that cannot be
>>classified to be Turing machines.
>>
>>>Which strongly suggests that a sufficiently butch computer could simulate
>>>a whole _lot_ of neurons, which could be networked together the same way
>>>as a brain.  I don't think it is considered _proved_ that the human
>>>brain is a Turing machine, but it is extremely strongly implied by all
>>>current research.
>>
>>I'm not buying that. In a real-life digital computer, quantum effects 
>
>I don't care if you buy it or not.  No one has yet proved neurons are
>_not_ Turing machines, and all the evidence in so far is that they _are_.
>You are arguing with axioms, not with logic, nor with evidence.
>
>>such as thermal noise have no effect on the course of the execution since
>>they are completely masked by the relatively large noise margin between
>>the two logic states. In the human brain which is mostly analog, noise
>                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
>Not only not proven, all evidence points back the other way.  While
>the brain is certainly not a classic von Neuman machine, most studies
>I've read show definite signs of an essentially digital device.  Again,
>you are not supported in this assertain by any major researcher I can
>think of.  More importantly, quantum effects have no effect on neurons,
>either.
>
>>in the neurons would have a cumulative effect via the so called butterfly
>>effect. If we run two simulations of the brain with the same initial 
>>state and input stimuli but with different sequences of random numbers
>>to model the neural noise, the two outcomes would diverge in no time.
>>Whether the variations are behaviorally significant is a moot point.
>
>That would be _so_ impressive, where it not possible to simulate chaotic
>behavior with a modern digital computer, which indicates that the nature
>of chaos _can_ be captured in a Turing machine.  I think you are ascribing
>limits to the computing ability of Turing machines because of your own
>prejudices.  You are appealing to some mystical animism, I'm afraid.
>
>While it is not _proven_, to my knowledge, that the human brain _is_ a
>Turing machine, all available evidence shows that it _is_.  You'll have
>to do better than this.
>
>-- 
>Larry Smith --- My opinions only.           larrys@zk3.dec.com/larrys@io.com.
>pentagon.io.com is Illuminati OnLine, SJ Games, _not_ "the" Pentagon, please.
>--
>"Wealth is crime enough to him that's poor." - Sir John Denham (1615-1669)

In article <3forh6$7ub@pentagon.io.com> larrys@pentagon.io.com (Larry Smith) writes:
>In article <vlsi_libD2oKoJ.GKw@netcom.com>,
>Gerard Malecki <vlsi_lib@netcom.com> wrote:
>>In article <3fmn70$cpa@pentagon.io.com> larrys@pentagon.io.com (Larry Smith) writes:
>>>
>>>Neurons have been simulated by a computer, which _is_ a Turing machine.
>>
>>This does not mean that neurons are themselves Turing machines. A neuron
>
>Err...yes, it does.  That's axiomatic, I'm afraid.  I don't care to go
>ito the proof here.
 
Simulations are only approximations of real phenomena. If the brain is
a Turing machine, then anything in the universe is a Turing machine (is
this the reason why some authors prefer the term 'Universal Turing machine?).
>
>>has a cretain degree of randomness to it, which may depend on quantum
>
>Turing machines are capable of sufficiently random behaviour to pass any
>test for it you care to name, even if the randomness _is_ algorithmic.
>The _source_ of randomness is utterly irrelevent for computation pur-
>poses, if the source can generate sufficiently random numbers.  Turning
>machines can be as random as needed, they can get asymptotically close
>to quantum variability.  You are looking for God under that curve, and
>He isn't there.
   
Oh yeah? Talk to any weather-modelling software developer. What may be
considered to be insignificant noise in the collected data would become
quite significant in a few days and completely overwhelm any kind of
prediction in about 10 days. And the brain is much more complicated than
global weather patterns.
 
>
>>phenomena. A main characterisitic feature of Turing machines is repeatability
>>which is not the case with real-life neurons. In fact, a lot of processes
>
>I won't speak for _your_ neurons, but neurons are repeatable within
>the design limits of their construction.  The simulation will not
>wear out or overload, as a real neuron can, but so long as that does
>not happen a neuron _is_ repeatable.  Bear in mind that neural nets
>exhibit a lot of what is now called "chaos", the patterns in networks
>are not directly repeatable, but are predictable and simulatable
>in principle, using chaos theory.
>
>>such as thermal noise in a resistor are of true random nature that cannot be
>>classified to be Turing machines.
>>
>>>Which strongly suggests that a sufficiently butch computer could simulate
>>>a whole _lot_ of neurons, which could be networked together the same way
>>>as a brain.  I don't think it is considered _proved_ that the human
>>>brain is a Turing machine, but it is extremely strongly implied by all
>>>current research.
>>
>>I'm not buying that. In a real-life digital computer, quantum effects 
>
>I don't care if you buy it or not.  No one has yet proved neurons are
>_not_ Turing machines, and all the evidence in so far is that they _are_.
>You are arguing with axioms, not with logic, nor with evidence.
 
What axiom are you talking about? Unless you can come up with a mathematical
proof that neurons are Turing machines, you cannot claim your assertion to
be an axiom.
>
>>such as thermal noise have no effect on the course of the execution since
>>they are completely masked by the relatively large noise margin between
>>the two logic states. In the human brain which is mostly analog, noise
>                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
>Not only not proven, all evidence points back the other way.  While
>the brain is certainly not a classic von Neuman machine, most studies
>I've read show definite signs of an essentially digital device.  Again,
>you are not supported in this assertain by any major researcher I can
>think of.
 
Give me a break! If your statement holds, people like Hertz, Krogh and
Palmer (from whose book I learnt most of my neural net stuff) should 
be 'minor' researchers. By the way, from where did _you_ learn anything about
brains or neural nets (assuming you learnt anything at all)? 
 
            More importantly, quantum effects have no effect on neurons,
>either.
 
A sweeping remark that hasn't been either proved or disproved so far.
 
>
>
>That would be _so_ impressive, where it not possible to simulate chaotic
>behavior with a modern digital computer, which indicates that the nature
>of chaos _can_ be captured in a Turing machine.  I think you are ascribing
>limits to the computing ability of Turing machines because of your own
>prejudices.  You are appealing to some mystical animism, I'm afraid.
 
Maybe, but it is much better than reaching premature conclusions.
>
>While it is not _proven_, to my knowledge, that the human brain _is_ a
>Turing machine, all available evidence shows that it _is_.  You'll have
>to do better than this.
>
You seem to be so authoritative in your views. Can you cite any references
to back your claims?
 
Shankar Ramakrishnan
shankar@vlibs.com
 
