From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!ccs-server.QueensU.CA!qucdn!kellydk Mon Dec  9 10:47:59 EST 1991
Article 1857 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!ccs-server.QueensU.CA!qucdn!kellydk
Organization: Queen's University at Kingston
Date: Wednesday, 4 Dec 1991 11:36:17 EST
>From: Doug Kelly <KELLYDK@QUCDN.QueensU.CA>
Message-ID: <91338.113617KELLYDK@QUCDN.QueensU.CA>
Newsgroups: comp.ai.philosophy
Subject: Re: Searle, again
References:  <2127@ucl-cs.uucp>

In article <2127@ucl-cs.uucp>, G.Joly@cs.ucl.ac.uk (Gordon Joly) says:
>
>From:    fb0m+@andrew.cmu.edu (Franklin Boyle)
>>> No, the "understanding" is in the heads of the people that wrote
>>> the book, and they're not in the book or the room.  Otherwise
>>> you would seem to be contradicting yourself by arguing Searle's point
>>> based on the relationship between language and social interaction.
>>> Thus, I assume you probably meant that understanding is reflected in
>>> the book's having been created.
>>>
>>> [...]
>>>
>>> Frank Boyle
>
>
>No, the understanding is in both the Book of Instruction and the
>Heads of the People who Wrote the Book of Instruction.
>
How is the understanding contained in the book?
Consider either of the two following two scenarios

1) I, being a potential member of the set of "People
who wrote the book of Instructions"  hand you a book of
symbolic rules IDENTICAL to the book in the chinese room.  The
only difference is that I made them up. As far as I, the creator of
the book, am concerned, the rules are jibberish.  Perhaps they
were derived from my cat walking across my keyboard.  Okay, now
I start feeding you some input symbols, to which you apply
the correct rules.  Thus, we have the same input symbols, same
output symbols, same system.  The only difference is that
NOBODY HAS ANY IDEA ABOUT WHAT IS GOING ON?!?!?!
How do you distinguish my jibberish book from the genuine
book that somehow mysteriously contains understanding.


2)I have a sequence of 1's and 0's (stored in a file, for
example)  What do they mean?


"Well, gee, heck- could be just about anything", I hear someone say.


Exactly.  This particular sequence of bits could be seen as
a valid input for any number of symbolic rule processors
similar to the chinese room.  (ie a sequence of 32 bits
could be
-4 characters
-an int
- a float
- a pointer to another location in  the file
.......etc.

Now, each of these symbolic manipulation packages is fed
the input, but only one produces the correct semantic output.
(the correct semantic output being defined by what my intent was
when I created the file - just like the 'correct' response
of the Chinese room is defined by the semantic content associated
with the symbolic input)


So, all of the manipulations are 'correct' in that they
faithfully apply the rules given to them, but only
one produces the correct output.  If I had created
the IDENTICAL stream of 1's and 0's through another
method, then a different package would be 'correct'.


thus, the idea of exhibiting understanding merely through
producing the correct output relies ENTIRELY upon
the assigned semantic meanings of the symbols.  This
occurs completely outside the scope of the system, and
so there can be no attribution of understanding to the
symbolic rules themselves.




In the above examples, either the semantic content, or the
symbolic content, but not both, was the same between
two or more systems.  I believe that this demonstrates
what I believe was Searle's point - a point, I must emphasize,
that I understand differently than others, it seems.

My reading of Searle is that the point he is trying to make
(in the Churchlands/Chinese room paper, at least)  is that

"A system can not possess understanding based solely on
symbolic maniuplation."


most of the arguements in this thread seem to miss this point.
Arguements about neural nets, robotic or analog systems with
physical sensors, etc.  fall entirely outside the systems that
Searle was considering in his arguement.  (If he addresses these
in subsequent work, my apologies)



Doug Kelly
kellydk@qucdn.queensu.ca


