Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!ames!amdahl!kp
From: kp@uts.amdahl.com (Ken Presting)
Newsgroups: comp.ai
Subject: Re: more Chinese Room
Summary: CR is about the information content of programs
Keywords: Chinese room, CR, Searle
Message-ID: <53tT02R288oV01@amdahl.uts.amdahl.com>
Date: 14 Feb 90 03:23:49 GMT
References: <1990Feb13.225830.13432@wam.umd.edu>
Reply-To: kp@amdahl.uts.amdahl.com (Ken Presting)
Distribution: usa
Organization: Amdahl Corporation, Sunnyvale CA
Lines: 94

In article <1990Feb13.225830.13432@wam.umd.edu> kohout@wam.umd.edu (Robert C. Kohout) writes:
>I must say that I am somewhat surprised to see this Searle discussion
>continuing. I haven't looked at this group in several months, and I 
>thought the Chinese Room had died a merciful death.

When Scientific American published Searle v Churchlands, we got going
again.

>        ...        One could summarize this view - I {Searle} call it
>    'strong artificial intelligence', or 'strong AI' - by saying that
>    the mind is to the brain, as the program is to the computer 
>    hardware."
>
>Do any of you actually agree with this summation? In particular, I want
>to point out that Searle equates mind with program. In all my time reading
>this group, I don't recall a single instance of such a statement. We may
>believe a lot of things, but do any AI practitioners/afficiandos out there
>actually think that mind/brain = program/hardware?

Sure, I think it's close enough.  Searle does seem to me to underestimate
the significance of *running* the program (:-).  If you count I/O devices
as hardware, then hardware and programs *better* be the terms of the
equation.

If I were to modify Searle's ratio, I would say (apologies to Wirth):

          mind/brain = (data structures)/(hardware+algorithm)

Searle talks almost exclusively about programs, seldom mentioning data.
It does bug me that Searle gives the CR no scratch paper.  It's just a
finite state machine.  But most of his argument doesn't depend on that,
or on the details of the analogy.


>2) Even given this notion of the relationship between mind and brain,
>which I regard more or less of a straw dog, Searle manages to bungle
>his argument. That is, later in the chapter, when he presents his now
>famous CR argument, Searle puts the human in the place of the hardware,
>and then points out that it is difficult to claim that such a person
>could understand Chinese. Remember, Searle claims to be trying to debunk
>the mind/brain = program/hardware notion. He posits a human who interprets
>a set of instructions about which he knows nothing. That is, he equates
>a human with the machine and the set of instructions with the program.
>Thus, by his own (albeit poor) analogy, one should not expect the human
>to understand, any more than one would expect the brain (vis a vis the Mind)
>to understand. Searle can't even shoot down his own weakly constructed
>creation.

If I follow you, you are saying that it's irrelevant that the person does
not understand Chinese.  And anyway, we didn't need an argument to tell
us that the person wouldn't understand - that's exactly what we would
expect based on the role the person has in the analogy.  After all, the
person is being compared to the Brain, and it's the Mind that understands.

The problem with this maneuver is that Searle is using his observation
(that the person doesn't understand) to make a very different point.
If you haven't seen the Scientific American article, you (may) want to
get a copy.  I think Searle's presentation there is very systematic and
well organized.

Searle uses the CR to support what he calls Axiom 3: "Syntax by itself
is neither constitutive of nor sufficient for semantics."  He goes on to
say "At one level this principle is true by definition."  So why does he
need to support it with the Chinese Room?  I'll get to that in a minute.

Suppose you program a Turing machine to accept some formal language, say
all the wff's in a first-order logic that use some finite set of
predicate letters and constants.  This TM program defines the syntax of
the language, and is entirely equivalent to some set of re-writing rules.
Now, neither the TM nor the equivalent grammar has anything at all to do
with the assignments of predicate letters to subsets in a model, nor the
assignments of constants to elements in a model.  That's Axiom 3.

The Chinese Room is like a Turing machine that doesn't just accept the
language.  After it accepts a wff, it prints out a new wff.  Searle's
point is that this TM is *also* completely independent of any semantics
for the language in question.  Of course, this sort of TM is what most
people have in mind for passing the Turing test, so Searle can conclude
that passing the Turing test does not require any knowledge of semantics,
that is, understanding.

Presumably Searle isn't up on automata theory, and had no better way to
state his idea than with a cute story.

>3) All Searle really shows is something that we all know already. No
>matter how grandiose our program, the bare metal of the digital computer
>as we currently know it will never itself become aware. Big deal. The
>Chinese Room might start a lot of great, go-nowhere discussions, but it
>proves, very, very little.

What it shows is that we'd better get clear on how computers are different
from TM's, or spruce up the Turing test, or both!

It certainly does *not* show that we have no choice but connectionism.