From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!psych.toronto.edu!michael Tue Mar 24 09:57:27 EST 1992
Article 4610 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!psych.toronto.edu!michael
>From: michael@psych.toronto.edu (Michael Gemar)
Subject: Re: The Systems Reply I
Organization: Department of Psychology, University of Toronto
References: <1992Mar6.185926.18497@oracorp.com> <1992Mar9.171606.6886@psych.toronto.edu> <1992Mar16.231755.32589@spss.com>
Message-ID: <1992Mar19.202259.16791@psych.toronto.edu>
Date: Thu, 19 Mar 1992 20:22:59 GMT

In article <1992Mar16.231755.32589@spss.com> markrose@spss.com (Mark Rosenfelder) writes:
>In article <1992Mar9.171606.6886@psych.toronto.edu> michael@psych.toronto.edu (Michael Gemar) writes:
>>The crucial move that Searle makes it to assume that if a mind (however
>>constituted) performs computations that would generate subjective
>>experience (understanding, qualia, whatever), then the mind performing
>>the computations should have these experiences.  
>
>And this assumption seems very strange to me.  It seems to underlie a number
>of statements you've made about the CR:
>
>"Even if the CR gave interpretable answers, the person doing the purely
>syntactic manipulations wouldn't understand."
>
>"If you do this [follow the CR rules], will you understand Chinese in the
>way you understand other languages?"
>
>"Because the issue at question is whether or not the CR process generates
>understanding in the *man*."

Well, I'm *darned* impressed that you've hung onto old postings, although it
it will certainly make me more cautious about what I say...

I should point out that the position that I have been trying to explicate
is the one that *Searle* puts forward.  I'm not so sure I buy his response
to the Systems Reply now, but I still think the above quotes are an accurate 
representation of his position.    

(Just as an aside, it's a mystery to me how I've become one of the primary
apologists for Searle.  I *certainly* don't believe his story about 
the brain's "causal powers"...)


>In the original CR story, the man surely does not understand Chinese; but
>what does this prove exactly?  In the metaphor, the man corresponds to the
>CPU, the rulebooks to the algorithm.  The story is thus a demonstration 
>that CPUs cannot understand.  But who maintains that they could?  
>If a computer is executing a program that passes the Turing Test, 
>it would be absurd to maintain that the computer itself is intelligent.
>Why, then, all this emphasis on whether the man in the CR, who plays the
>role of the CPU, understands or not?  
>
>The situation doesn't really change by having the man memorize the CR
>rules.  His conscious mind still performs the role, merely, of 
>unintelligent executor of the CR algorithm.  There is no reason to
>expect that he will have any insight into the algorithm or any
>understanding of Chinese, any more (to use another poster's example)
>than a Macintosh executing Soft-PC itself runs DOS programs.

As I note above, I'm beginning to think you're right on this score.
Searle's response to the Systems Reply does not get at the heart of the
move the Systems Reply makes.  I *still* think that the problem of generating
meaning from symbol-shuffling remains, however...


>All this said, I think the CR story helps point out a difference between
>human and (traditional symbolic) computer operation.   Why does the 
>man in the CR, or the CPU, not understand Chinese?  Surely part of the
>answer is that they are such narrow informational channels.  The CPU
>can handle only a few floating point numbers at a time; the man, only
>a few simple rules.  Whatever mental phenomena occur in the system 
>must occur at a higher level, which cannot fit into this narrow channel.
>Contrast the brain, which has a billion little processors running
>concurrently; here we run into no such width limitations.  

This point is not at all clear to me, since we assume that, in the CR, the
operations which are producing the mind occur *serially*.  In general, 
*any* parallel architecture can be converted to serial (almost *all*
connectionist work is done on conventional, serial machines).

- michael



