From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!samsung!emory!gwinnett!depsych!rc Tue Jan 21 09:27:22 EST 1992
Article 2911 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!samsung!emory!gwinnett!depsych!rc
>From: rc@depsych.Gwinnett.COM (Richard Carlson)
Newsgroups: comp.ai.philosophy
Subject: Re: Searle Agrees with Strong AI?
Message-ID: <V0sueB3w164w@depsych.Gwinnett.COM>
Date: 20 Jan 92 15:09:06 GMT
References: <1992Jan19.213507.11148@bronze.ucs.indiana.edu>
Lines: 108

chalmers@bronze.ucs.indiana.edu (David Chalmers) writes:

> In article <o2wseB3w164w@depsych.Gwinnett.COM> rc@depsych.Gwinnett.COM (Richa
> 
> >Everyone seems to be ignoring the fact that the program
> >(presumably in an improved and more powerful version) which Searle
> >assumed was being implemented by the Chinese Room was itself
> >essentially an indexed lookup table, Roger Schank's table of
> >scripts -- in particular Searle discussed a restaurant script.
> >
> >Since Schank suggests, quite plausibly, that much of our daily,
> >regularized, mundane behavior -- until some interrupt occurs, and
> >often even then -- is in fact taken from such a lookup table, it
> >really isn't "cheating" for a computer to use one.
> 
> 1. Nothing essential to Searle's argument rides on the program's being
> Schankian, as Searle himself makes clear.
> 
> 2. Even Schank's "look-up tables" are far, far more sophisticated than
> the brute-force mechanism described earlier.

I don't think it makes any _logical_ difference what the program
in the Chinese Room happens to be, but I think the fact that it
was a program for the interpretation of narrative strengthened
Searle's argument in a psychological and rhetorical sense.

I haven't read Searle's original writings on the Chinese Room.  I
got most of what I know from Penrose's discussion in _The
Emperor's New Mind_, a reference I got from this Newsgroup, and
from postings in this Newsgroup.  Penrose says that Searle chose
Schank's indexed script program since it _already_ passed a
limited Turing Test.  You could ask the computer (or the Chinese
Room) whether or not a patron had eaten a hamburger and "it"
(referring to the "system," whether a computer running Schank's
software, including his indexed scripts for restaurant behavior,
or a person following the instructions contained in the program,
but now translated into English, in a rote manner) would give you
a correct answer -- yes, he ate the hamburger, or no, he didn't
eat the hamburger.

Although it took a while, everybody now seems agreed that the
Chinese Room reduces to the Turing Test, with the exception that
now there is a human being in the loop, although not using any of
his human abilities except for those that would be involved in
carrying out simple instructions.  Basically Aristotle could have
run a form of the Turing Test to prove that with his analytics,
which was probably the first stab at some kind of automated, or
"artificial" thinking, an ordinary clerk, without special
reasoning ability. could arrive at the same true conclusions as
the wisest philosopher merely by manipulating symbols on a piece
of parchment (or whatever it was that the old Greeks wrote on).
Such a test, if it had occurred to Aristotle to run one, would
have treated the fact that the ordinary clerk did not "understand"
the "thinking" involved in coming to the true conclusion as proof
_for_ the existence of "strong AI," since the key thing for him
and his audience would be that the "thought" had been moved
_outside_ the mind of the wise philosopher that the "humanists" or
"vitalists" of the day would probably have maintained was
necessary for thought to occur.  The fact that parts of the
procedure involved activities going on in the mind of the clerk
would have seemed irrelevant to them.  It would probably have
reminded them of the very ordinary circumstance of an intelligent
master telling his servant to begin planting on a certain day,
although the servant would not be expected to understand the
reasoning as to why that was the right day to do it -- all the
servant had to do was carry out the master's orders, using his
hands and back, directed, as it were, by the master's brain.

The fact that there is a human clerk, presumably with a high IQ
since Searle volunteers himself for the (thought) experiment,
although working, if not from the neck down, at least without
using any higher cortical powers beyond following simple
instructions, has exactly the _opposite_ effect today.  It seems
to detract from the claims of "strong AI."  But should it?  What
is the functional significance of putting a human being into a
loop in which he performs essentially non-human functions? Suppose
a fantastically brilliant Chinese peasant was pulling a rickshaw
down the streets of Hong Kong on January 20, 1892. The two English
intellectuals riding in the rickshaw were discussing the mental
capacities of the Chinese and one of them gave as an example the
observation that this Chinese peasant had carried him several
times to a certain location but had never learned the streets by
rote and always had to pause to recall a turn, so, the first
English intellectual concluded, this Chinese peasant isn't even as
intelligent as the average horse, who can perform such feats of
memorization after making the trip a number of times.  Maybe his
companion accuses him of naivete and asserts that the Chinese
really knows the route but hesitates and loiters out of sheer
laziness. Meanwhile, unbeknownst to them this brilliant man is
composing in his head the most intricate axiom system for a kind
of mathematics which wasn't even dreamed of then but which maps
well onto some contemporary need and will be discovered on January
21, 1992, in an old journal the brilliant peasant dictated to a
scribe who believed in his brilliance.  Within the rickshaw system
this genius was nothing but a beast of burden.  Within the Chinese
Room Searle (or some luckless graduate student!) is nothing but a
clerk.  (Maybe Searle or the graduate student will think of some
brilliant addiction to speech act theory while carrying out his
mindless task of copying symbols.)  What possible significance
does that have, even in principle, to a paradigm, the Turing Test,
which nearly everyone has now concluded, has no definitive
scientific significance?

--
Richard Carlson        |    rc@depsych.gwinnett.COM
Midtown Medical Center |    {rutgers,ogicse,gatech}!emory!gwinnett!depsych!rc
Atlanta, Georgia       |
(404) 881-6877         |


