From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!convex!linac!mp.cs.niu.edu!rickert Tue May 12 15:49:52 EDT 1992
Article 5500 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!convex!linac!mp.cs.niu.edu!rickert
>From: rickert@mp.cs.niu.edu (Neil Rickert)
Subject: Re: Games (was Re: Categories: bounded or graded?)
Message-ID: <1992May8.222740.2226@mp.cs.niu.edu>
Organization: Northern Illinois University
References: <6649@skye.ed.ac.uk> <OZ.92May5014616@ursa.sis.yorku.ca> <6690@skye.ed.ac.uk>
Date: Fri, 8 May 1992 22:27:40 GMT
Lines: 63

In article <6690@skye.ed.ac.uk> jeff@aiai.ed.ac.uk (Jeff Dalton) writes:
>In article <OZ.92May5014616@ursa.sis.yorku.ca> oz@ursa.sis.yorku.ca (Ozan Yigit) writes:
>>jeff@aiai.ed.ac.uk (Jeff Dalton) writes:
>>
>>
>>   		... It's very difficult to come up with
>>   definitions that can't be "messed up", but that doesn't mean
>>   we never know what we're talking about.
>>
>>But that also doesn't mean we do know even if it appears so, which
>>as you so conveniently ignore, is the exact reason why Searle gets
>>so much milage out of his miraculous gedankenexperiment.
>
>Of course, people who disagree with Searle would like it to be
>the case that no one has good reasons for agreeing with him.

  I have never argued that there are not good reasons for agreeing with
Searle.  I am every bit as skeptical as Searle about attempts at symbolic
AI, for example, and for quite similar reasons (such as the relation between
syntax and semantics).  I agree that the syntax/semantics argument does
much to capture why many experts systems, for example, fail to duplicate the
power of the mind.  But ultimately I must reject Searle's argument, for his
conception of the capabilities of a computer is just far too restricted.
The "good reasons" for agreeing with Searle evaporate once you realize how
flexible and extensible computers really are.

>But they really ought to try to show that Searle gets so much
>milage for bad reasons, rather than merely asserting it.

  The bad reason Searle seems to succeed is that he takes a precise
formal definition of a computer (as a Turing Machine), and then uses it
in a very informal and imprecise argument.  Formal definitions are great
for formal arguments, but they usually make little sense in an informal
discussion.  But he doesn't prove anything - he just makes AI sound absurd.

  Try these:
		A stereo system is an electrical apparatus for
		wiggling a piece of paper.

	This makes it sound absurd that you could ever use it to
	reproduce music.

		I can count my fingers by putting them in a one to one
		association with the elements of a set of ordered pairs of 
		Dedekind cuts of equivalence classes of equivalence
		classes of sets of sets.

	This makes it sound as if you would never be able to count.

  The effect of using a formal treatment of computers in the intuitive CR
argument is to paint a picture of computers which is quite absurd.  For
example it sounds absurd to think that by syntactically manipulating
symbols you can construct detailed models of the atmosphere to study the
development of such as tornados or hurricanes or El Nino, phenomena which
have not been amenable to direct mathematical analysis.

  Of course you could try to correct the problem by giving precise formal
definitions of such terms as "understanding" and "semantics", and using
these in a formal mathematical proof.  Or you could try to correct the
problem by using a more reasonable intuitive description of a computer
which clearly encompasses all the things that we know can be done by
computers.



