From newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!mp.cs.niu.edu!rickert Mon Jun 15 16:04:37 EDT 1992
Article 6201 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!mp.cs.niu.edu!rickert
>From: rickert@mp.cs.niu.edu (Neil Rickert)
Newsgroups: comp.ai.philosophy
Subject: Re: Transducers
Message-ID: <4138.708217481@mp.cs.niu.edu>
Date: 10 Jun 92 23:04:41 GMT
References: <1992Jun10.203412.19158@news.Hawaii.Edu>
Organization: Northern Illinois University
Lines: 89

In article <1992Jun10.203412.19158@news.Hawaii.Edu> roitblat@uhunix.uhcc.Hawaii.Edu (Herbert Roitblat) writes:
>
>     I would like to contribute to the "I am transducer"
>argument.

  Your thoughtful contribution (most of which I have not included in this
response) is appreciated.

>     An alternative to the mind/transducer argument suggests that
>the mind is a central computational core surrounded by storage
>and peripheral devices.

  The word "alternative" is at the heart of what I consider to be
a serious misunderstanding.

  We are looking at two possibilities:

	1: The mind is a transducer

	2: The mind is a set of peripherals connected to a computation
	   core.

  So far, so good.  But you and Harnad both point to various points of
evidence and essentially say "Look at this evidence - it must be a
transducer.  Possibility 1 is true, so possibility 2 is false."

  But I believe this is a wrong approach.  We should consider that both
possibilities could equally be true.  That is, they could both be
descriptions of the same system.  The neurobiologist needs to concentrate
mostly on possibility 1, because that highlights his approach to the
subject.  The computer scientist needs to concentrate on possibility 2,
which highlights his approach.  Perhaps the psychologist needs to be
able to switch back and forth between the two views.

  To put this in perspective, consider an automobile.  This doesn't look
like a set of peripherals and a computational core.  But it can still
be described that way.  The computational core view emphasizes the
equations which relate the input (pressure on the brake, position of the
gas pedal, etc) to the outputs (frictional forces on the brake drums,
engine thrust, etc).  It doesn't much matter whether the automobile
physically seems to fit the computational view.  It is still a useful
view which allows you to ignore details such as whether the gear
lever is floor mounted or steering column mounted.  Interestingly,
as automobiles and aircraft become ever more sophisticated, the
physical description seems to approach the computational
description.

 In summary, there is much in Harnad's view and your view of
cognition which is thoughtful and suggestive.  My one objection
is when you claim that it is exclusive of the computational
view.

>                                             A major portion of
>the brain is intimately involved in perception and motor control
>and these are the same portions that seem to be required for
>intelligence.  In support of such a claim that perception and
>motor control are the seats of intelligence, psychological
>research finds that intelligent performance is often the product
>of pattern recognition rather than superior computational skill
>(e.g., de Groot, 1966).

  I certainly agree that pattern recognition is important.  If
anything, I would claim that it is far more important than most people
are willing to admit.  But why does pattern recognition rule out
computational skill?  Surely pattern recognition is a computational
task of enormous complexity.

  Much of discussion about AI seems to reach a deadlock at exactly the
point of the nature of computation.  You shouldn't think of computation as
just plugging some numbers into a formula and getting out a result.  You
can have computation which uses statistical methods to sort through large
masses of data, detecting frequently recurrent patterns.  Once the patterns
have been detected, their future reappearance can in principle be recognized.
This pattern detection (i.e. learning), and subsequent pattern recognition
surely constitute a computational task.  The computational problem is not
well understood at present.  Once it is understood, we may be well on our
way to resolving the fundamental problems of AI.

  I agree with Harnad's assertion that thought (or at least human
thought) is physical.  I can find considerable evidence to support
that view.  I don't see the evidence that pattern recognition is
physical and non-computational.  Indeed, pattern recognition seems to
be too rapid to have much of a non-computational component.  The
learning, or pattern detection, may be a different matter.  In the human
brain it is plausibly the result of physical growth, and if so could
be considered physical.  But growth may well be a biological way of
performing statistical procedures.  Only by either better biological
knowedge about learning, or by the production of adequate computational
models of learning, will this finally be settled.


