Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uunet!mcsun!hp4nl!kubix!weigand
From: weigand@kub.nl (Hans Weigand)
Newsgroups: comp.ai
Subject: Re: Can machines think....
Message-ID: <773@kubix.kub.nl>
Date: 23 Feb 90 14:28:03 GMT
References: <2313@ritcsh.cs.rit.edu> <1990Feb19.165835.9673@pcsbst.pcs.com> <3964@cbnewsj.ATT.COM>
Reply-To: weigand@kubix.UUCP (Hans Weigand)
Organization: KUB Univ., Tilburg, The Netherlands.
Lines: 45

In article <3964@cbnewsj.ATT.COM> jwi@cbnewsj.ATT.COM (Jim Winer @ AT&T, Middletown, NJ) writes:
>The answers that we provide to the questions:
>	Can submarines swim?
>and
>	Can airplanes fly?
>may tell us a lot about the subject of whether machines can think.
>
>In paritcular, many of us would concede that airplanes fly, and some of 
>us that submarines swim for all practical purposes. By extension then,
>machines that perform a practical function that normally requires
>thinking can be said to think. Let us keep in mind that our answers to
>all three of these questions should be reasonally analogous.

This reasoning by analogy has its pitfalls. The question is of course
whether the terms (swim,fly,think) are functional or not (that is,
can be described exhaustively in functional terms). There is 
already a difference between the swimming and the flying. The latter
is more accepted in English, because flying means roughly something
like "MOVE IN AIR ON OWN FORCE", whereas swimming is more than "MOVE
IN WATER ON OWN FORCE": it implies a certain kind of moving, inside
or on top of the water, that we attribute to humans, fishes, ducks,
but not to ships, submarines, and surfers. Redefining the meaning
of "to swim" to a purely functional term is no solution of course:
then you can say that a submarine can swim, but you still have not
characterized the swimming in its original sense.

(A similar problem exists already for characterizing "flying" with
respect to "move in the air". A jumping frog does not fly).

When we go to the term "thinking", again we have to ask ourselves
whether the functional performance completely captures the concept
or not. The fact that non-biased humans tend to accept the CR argument,
suggests that this is not the case. Redefining the term, as computer
scientists want to do, is again no real solution, although it makes your
view of the world a lot simpler perhaps.

I see no a priori reason why "the three answers should be reasonably
analogous" unless one has already chosen for a dogmatic functionalism.
For me, some terms may have a functional definition, some terms depend
just on linguistic conventions, and some others may be beyond the grasp
of both functionalism and conventionalism. Sometimes the world is a
bit more complicated then we would like ...

--- 
Hans Weigand