From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!watserv1!watdragon!logos.waterloo.edu!cpshelle Thu Apr 16 11:33:54 EDT 1992
Article 5032 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!watserv1!watdragon!logos.waterloo.edu!cpshelle
>From: cpshelle@logos.waterloo.edu (cameron shelley)
Subject: Re: syntax and semantics
Message-ID: <1992Apr10.134716.26504@watdragon.waterloo.edu>
Sender: news@watdragon.waterloo.edu (USENET News System)
Organization: University of Waterloo
References: <1992Apr9.204735.21732@psych.toronto.edu>
Date: Fri, 10 Apr 1992 13:47:16 GMT
Lines: 68

michael@psych.toronto.edu (Michael Gemar) writes:
> In article <1992Apr8.215800.18021@mp.cs.niu.edu> rickert@mp.cs.niu.edu (Neil Rickert) writes:
[...]
> > It is of course possible to view floating point arithmetic as formal and
> >precise.  If you view it that way the floating point numbers satisfy some
> >very strange and quite complex algebraic properties.  Moreover the
> >algebra of floating point numbers as implemented on one machine is quite
> >different from the algebra as implemented on a different machine.  All in
> >all, when viewed this way floating point numbers are useless curiosities.
> >However if view as approximations to real numbers then we have a much
> >simpler algebra (the algebra of real numbers), we have consistency between
> >different machines, but the numbers are no longer precise, since they are
> >approximations.
> 
> This is *not* the issue.  The computation *is* formal, and *is* precise.  The
> fact that the result you get does not always jibe from machine to machine
> is *not* to say that the computations themselves are not precise and
> formal.

Actually, I think Neil is correct in his conclusion, although I don't
think floating point arithmetic is a good example.  It is *programs*
that are formal systems.  *Computers* are pieces of hardware subject
to unanticipated vagaries of performance.  Floating point operations
are certainly a problem, but more common are the effects of race
conditions, resource deadlocks, I/O errors, and the like.  Computer
design is not just an issue of making the result equivalent to a UTM,
as Searle would have us think.  

While Searle, and I think Michael, restrict discussion to programs and
hardware, the indispensible issue of *process* behaviour gets ignored.
Actual instantiations of programs on real computers (not the paper
variety) are what's at stake in the "strong AI" program.  This is very
easy to forget if you are isolated from such problems.  Searle does
this to great effect in the Chinese room argument: applying the terms
"syntax" and "semantics" identically to programs, processes, and
natural languages.  He never offers any justification of we can make
such an assumption.

Searle is quite correct in most of his argument.  Programs are formal
systems.  Technology is not at issue.  A program, running on an ideal
computing platform (whatever that is) will never be more than a
mapping of a formal system onto a (completely isolated) physical
object.  Einstein spent years filling thought-boxes with springs,
pulleys, and clocks, but never did manage to shut out the universe. 
Before anyone posts about virtual realities, let me just remark that
if you give me a formal symbol system embedded in n superordinate
symbol systems, I will give you a single symbol system which computes
the same results.  (This is, of course, a rhetorical offer...)  If you
maintain that only the properties of formal systems are at issue, then
this is perfectly acceptable.

To address Michael's point about the formal nature of computations:
have you ever programmed a natural language interface, or some other
program that must accept input and operate under conditions that it
cannot fully anticipate?  Write software that is tolerant of hardware
failure?  Ultimately, if we suppose that intelligence is a property of
programs and hardware under stressless conditions, then the claim is
that intelligence is a property of a symbolic system.  I, speaking for
myself, don't accept this.  If we suppose, however, that intelligence
is a property of a situated process, then Searle's argument is
inconclusive, since it is situated under `thought' conditions.

				Cam
--
      Cameron Shelley        | "Proof, n.  Evidence having a shade more of
cpshelle@logos.waterloo.edu  |  plausibility than of unlikelyhood.  The
    Davis Centre Rm 2136     |  testimony of two credible witnesses as
 Phone (519) 885-1211 x3390  |	opposed to that of one."    Ambrose Bierce


