From newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!ux1.cso.uiuc.edu!mp.cs.niu.edu!rickert Mon Jun 15 16:04:54 EDT 1992
Article 6228 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!ux1.cso.uiuc.edu!mp.cs.niu.edu!rickert
>From: rickert@mp.cs.niu.edu (Neil Rickert)
Subject: Re: Re: Transducers
Message-ID: <1992Jun12.172225.27739@mp.cs.niu.edu>
Organization: Northern Illinois University
References: <1992Jun10.203412.19158@news.Hawaii.Edu> <4138.708217481@mp.cs.niu.edu> <1992Jun12.022620.22946@news.Hawaii.Edu>
Date: Fri, 12 Jun 1992 17:22:25 GMT
Lines: 180

In article <1992Jun12.022620.22946@news.Hawaii.Edu> roitblat@uhunix.uhcc.Hawaii.Edu (Herbert Roitblat) writes:
>In article <4138.708217481@mp.cs.niu.edu> rickert@mp.cs.niu.edu (Neil Rickert) wrote:

>>	1: The mind is a transducer
>
>>	2: The mind is a set of peripherals connected to a computation
>>	   core.

>>                                           We should consider that both
>>possibilities could equally be true.  That is, they could both be
>>descriptions of the same system.  

>If these hypotheses are mutually exclusive then they cannot both be
>true simultaneoulsy.

 After I explain why the two hypotheses are not mutually exclusive, you
seem to imply that I said the exact opposite.

>                                          Contrary to what you seem
>top be saying, the data do matter.  Freed of the constraints of data,
>any description can be attached to any object.

  I totally agree.  The data matters.  So, let's look at the data.

>>  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.  
>
>An automobile is about as far from a computational core device as one
>can get.

  I deliberately chose this example because superficially it seems so far
from the computational view.

>                       If the car had a computational core, a signal
>would travel from the brake to the core and from the core to the
>device.

  It does!  It does!  You just can't see it, because you are so convinced
it is not there that you aren't even looking for it.

>         Attempts to repair a car based on such a computational core
>theory would surely end in failure and possibly in death.  

  Attempts to repair the computer on your desk based on a computational
core theory would also surely end in failure.  The computer repairman
had better repair the physical components, and not even think about
bits and bytes, algorithms and computation.

  Look more closely at the automobile.  This time, let's look at the
real data, and not be confused by external appearances.

  The brake consists of:

    An input peripheral, to which the operator applies a small force.  This
    is typically the brake pedal.

    An output peripheral which generates a large force to resist motion.
    This commonly consists of the brake shoes being pushed against the
    brake drums to create a frictional force.

    A computational core which computes the output force from the input
    force.  The computation is rather simple -- it amounts to 
    multiplication by a constant.  It does not require high precision --
    a 25% error would probably not be easily noticed.

    The computation can be performed in many ways.  In the oldest cars,
    it was performed by using simple property of levers.  For many
    decades it has instead been performed by hydraulic systems.  In
    many of today's automobiles it is partly performed by the power
    brake system.

  Now let's look at it as a physical system, and not a computational
system.  We need to find the essential physical components.  What are
they?

    Is the lever important?  No, it can't be, or you wouldn't be able
    to replace the lever by a hydraulic system.

    Is the hydraulic system important?  It can't be, or the air brakes
    on large trucks would not work.

    Is it that a mechanical system uses only the energy supplied by
    the operator?  No, that can't be important, or we would not have
    the power brake systems.  You could not stop a large truck or
    a train with only the force supplied by the operator. 

    Is the use of brake shoes and drums the critical physical aspect?
    No, that can't be, since many cars use disk brakes instead.

    Is it the use of friction?  No, because that would not explain the
    braking of aircraft by reverse-thrusting the jet engines.

    Is it the use of a pedal?  That doesn't seem right, either, since in
    trains a hand control is used to apply the brakes.

 We've looked at the data.  There is not a single physical aspect which
essential.  But the peripheral/computational view holds up under all
variations of braking systems.

>>  I certainly agree that pattern recognition is important.  . . .

>> But why does pattern recognition rule out computational skill?

>Here you seem to have misunderstood the nature of the claim.  Master
>chess players are not necessarily smarter than poorer players, and
>they do not appear to employ better search heuristics.  Rather, they
>seem to be able to recognize more sensible chess positions than poorer
>players recognize and they know what moves have worked in the past in
>the presence of those board configurations.  Search heuristics are
>computationally very expensive.  If one wanted to MODEL human chess
>masters (as opposed to building successful chess-playing programs),
>the way to do it would be with improved pattern recognition and
>storage algorithms rather than with improved tree-traversing
>algorithms.  

  Once again you take my statement, and interpret as meaning the exact
opposite of what I said.  After I very clearly stated that pattern
recognition was important, you accuse me of ignoring the importance of
pattern recognition, and depending instead on tree traversing algorithms.
Yet I never once mentioned tree traversing algorithms.  I never once
suggested that I thought heuristic searchs were the secret.

  What is it with you?  Is this a deliberatele malicious attack,
or is it a simple matter of your own computational core failing you?

>>Surely pattern recognition is a computational task of enormous
>>complexity.

>Surely you are correct.  Nevertheless being committed to computation
>does not commit one to a certain computational architecture.

  The only person in this discussion committed to a "certain computational
architecture" is you.  You are inventing an architecture which you
believe to be inappropriate, and are using your invention to accuse
me, without one iota of evidence that I am committed to that
architecture.

>                                                              You seem
>to conflating computation with the idea of a computational core.

  The term "computational core" was introduced by Harnad and by you.  If
you don't like the term, you don't have to use it.  But don't toss it up
as a strawman just to give yourself a bogus argument.

>                                                                  The
>idea of mind as transducer implies that the computations are
>distributed throughout the system and cannot be concentrated in a
>disembodied core onto which the peripherals project their sensory
>information.

  Let's get our facts straight.

  Firstly there is no computation going on in the brain.  There are just
lots of chemical and electrochemical reactions.

  Secondly, there is no computation going on in the computer I am using.
There are just electrons moving around, quartz crystals vibrating, etc.

  The computation does not exist in the computer or the brain.  The
computation exists in the mathematical model we create to describe the
actions of the computer and the brain.  There is no single locus for the
computational core in the brain.  There is no single locus for the
computational core in the computer.  The locus of the computational core
is in the mathematical models we create to describe what we consider to
be the important aspects of what happens in the computer. The computational
core of the mind, when we reach that point, will be in the mathematical
model we build to explain how the brain creates intelligence.

  In a modern computer, when you look at a single number defined as a
single word of memory, that number is likely to consist of one bit
in each of 32 separate memory chips.  The physical arrangement is
nothing at all like the single memory location in our formal
descriptions of computing.  Computation in a computer is highly
distributed amongst the physical components of that computer.  In other
words, your strawman computational core model has almost nothing to do
with the actual physical organization of the computer.  So why should
the fact that it has nothing to do with the physical organization of
the brain be at all remarkable?



