Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!emory!mephisto!mcnc!rti!ntpdvp1!kenp
From: kenp@ntpdvp1.UUCP (Ken Presting)
Newsgroups: comp.ai
Subject: Searle and Radical Translation (was: Re: Searle and Biology)
Summary: Searle cares only about semantics, which is objectively testable
Message-ID: <606@ntpdvp1.UUCP>
Date: 18 Jul 90 19:16:39 GMT
References: <14265@enera.isi.edu> <602@ntpdvp1.UUCP> <1595@oravax.UUCP>
Organization: SNA Solutions Inc., Contract Programming Group
Lines: 60

In article <1595@oravax.UUCP>, daryl@oravax.UUCP (Steven Daryl McCullough) writes:
> In article <602@ntpdvp1.UUCP>, kenp@ntpdvp1.UUCP (Ken Presting) writes:
> 
> > It is this: We know for a fact that brains think.  We don't know at
> > all whether anything else will ever think.  Nobody in his right mind
> > would deny the first assertion.  That is the extent of the
> > "superiority of biological implementations."
> 
> . . .  I would hesitate to say that "we know for a fact that brains
> think" until we agree on what counts for knowledge in something like
> this. . . .

Sorry to have been obscure - I was hoping merely to put to rest an 
unproductive issue.

By "Brains Think" I of course meant "Brains Can Think".  Anyone who denies
this thereby implies either that he is not thinking (which is absurd), or 
that he is thinking with something other than his brain (his organs of 
succession, perhaps).

There are, of course, people with brains that can think, but instead of
using the capacity to think for themselves, merely quote or misquote the 
thoughts of others.
> 
> However, Searle denies the validity of using behavior as a test for
> intelligence. What else is there? Well, introspection: I know that I
> think because I experience myself thinking. This introspection doesn't
> get me any closer to knowing that *brains* (plural) think; only that
> *I* think.

This is either a misreading of Searle, or an (unwarranted) assumption about
the implications of his views.  He is concerned *only* with programs
and the consequenceds of implementing them.  In particular, he denies
that implementing a program attaches any semantics to the symbols it 
prints.

Searle does not say much about how to determine whether symbols have any
semantics attached, but he is unlikely to object strongly to the "Radical
Interpretation" of Davidson (and Quine), which is well-known and very 
influential among philsophers of language.

Radical Interpretation *is* based on observations of behavior.  (Cf. WVO
Quine, _Word and Object_, and D. Davidson, _Inquiries into Truth and 
Interpretation_)

Radical Translation involves adopting the hypothesis that the symbols
generated by a system are *true sentences* (among other assumptions), and
iteratively modifying a tentative semantics for the system, hoping to 
preserve the hypothesis.  The interpretation "succeeds" if some symbol
in the system's vocabulary gets assigned the semantics of the concept 
"truth".  (This is a very quick sketch of a very complex and interesting
theory, which was never intended to be applied to AI.)

Note that no mere implementation of a Turing Machine could pass this test,
because truth (like "halting") cannot be represented by a computable
function.  IMO, real computers are more than mere TM implementations, so
Tarksi's theorem does not prohibit real machines from representing truth.


Ken Presting