From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!yale.edu!qt.cs.utexas.edu!cs.utexas.edu!uunet!mcsun!uknet!slxsys!warwick!nott-cs!ucl-cs!news Mon Dec  9 10:47:56 EST 1991
Article 1851 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!yale.edu!qt.cs.utexas.edu!cs.utexas.edu!uunet!mcsun!uknet!slxsys!warwick!nott-cs!ucl-cs!news
>From: G.Joly@cs.ucl.ac.uk (Gordon Joly)
Newsgroups: comp.ai.philosophy
Subject: Re: Neural nets and the Chinese room
Message-ID: <2137@ucl-cs.uucp>
Date: 4 Dec 91 15:50:58 GMT
Sender: news@cs.ucl.ac.uk
Lines: 56

Martin Taylor writes:
 > 
 > > It took listening to Pribram for me
 > >to realize that Penrose has some real points. Although quantum processes
 > >are probably too low a level to have significant influence on brain
 > >function, the mathematics that appears to apply to the description of
 > >brain dynamics is the same statistical dynamics used to analyze quantum
 > >systems.
 > >
 > >-- 
 > >Harry Erwin
 > >Internet: erwin@trwacs.fp.trw.com
 > 
 > Funny you should say that, because it was precisely the dynamic view of
 > brain function that led me to believe that Penrose's arguments were
 > irrelevant, and to say so in my contribution to the multiple book
 > review of Penrose in Behavioural and Brain Sciences.  Penrose deals
 > in algorithms, generalized as processes that, given a set of initial
 > data will produce a determined result, and that cannot be disturbed
 > until they are completed.  Intermediate states in the execution of the
 > algorithm, according to Penrose, can tell you nothing about the final
 > result.  It's a temporarily isolated black box.  But everything in the
 > brain is subject to disturbance, and nothing algorithmic (in that sense)
 > could possibly operate in a physical brain.  Issues of time-scale are
 > important here, because disturbances that occur on a time-scale much
 > longer than the probable execution of an algorithm are unimportant; the
 > algorithm can be adapted (we don't concern ourselves with continental
 > drift when consulting TransAtlantic airline schedules).  Matters that
 > affect our conscious behaviour occur on a time-scale of hundreds of
 > milliseconds to weeks or years.  They don't disturb neural events that
 > are essentially complete in a few msec (much), so it has seemed reasonable
 > to talk about neural systems as if they could execute algorithms.  But
 > there's no need to go to the ends of the universe to explain why that
 > isn't strictly true.
 > -- 
 > Martin Taylor (mmt@ben.dciem.dnd.ca ...!uunet!dciem!mmt) (416) 635-2048
 > Beauty being skin deep, it would seem that the hippopotamus has an
 > epidermal advantage that it has somehow failed to exploit. (Jack Kent)



What should the quantum level not be important to the brain?  Quantum
effects can be writ large: for example the Zenner Diode.

[ Researchers in Neural Networks obviously think that the neuron is
the correct granularity. ]

When I spoke to Penrose, about 18 months ago, he did had not heard of
genetic algorithms.
____

Gordon Joly                                       +44 71 387 7050 ext 3716
Internet: G.Joly@cs.ucl.ac.uk          UUCP: ...!{uunet,ukc}!ucl-cs!G.Joly
Computer Science, University College London, Gower Street, LONDON WC1E 6BT

  he said what? - he said "hello" -- that's a bit aggressive isn't it?


