From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!batcomputer!cornell!rochester!yamauchi Sun Dec  1 13:06:46 EST 1991
Article 1765 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!batcomputer!cornell!rochester!yamauchi
>From: yamauchi@cs.rochester.edu (Brian Yamauchi)
Subject: Re: A Behaviorist Approach to AI Philosophy
In-Reply-To: Franklin Boyle's message of Fri, 29 Nov 1991 17:16:18 -0500 
Message-ID: <YAMAUCHI.91Nov30002306@magenta.cs.rochester.edu>
Sender: yamauchi@cs.rochester.edu (Brian Yamauchi)
Nntp-Posting-Host: magenta.cs.rochester.edu
Organization: University of Rochester
References: <AdBfkmC00WBME1JqUw@andrew.cmu.edu>
Date: 30 Nov 91 00:23:06

In article <AdBfkmC00WBME1JqUw@andrew.cmu.edu> Franklin Boyle <fb0m+@andrew.cmu.edu> writes:
>Brian Yamauchi writes:

>> True, but as many, many people have pointed out -- Searle isn't right.
>> I still have yet to see a convincing rebuttal to the "Systems Reply".

>Well, let me try the "Nonsense Room", which I briefly outlined at
>the end of a previous posting.

>Suppose we have what might be called the "Nonsense Room", which is similar
>to the Chinese Room except that the symbols fed into the room and those
>which form the rules in the rule book for processing the input (that is, 
>the part of the rule book which is not part of the instructions on how to 
>use it) were intended by the creators of that book to refer to 
>nothing. That is, they were never intended to have any meaning whatsoever!  

>So, the only difference between these two situations is that the Chinese
>symbols have referents for someone who "knows" Chinese.  But these
>referents are unknown to the person in the Chinese room *and* to the room + 
>person + book, etc.  All of these entities "see" only symbols whose forms are 
>arbitrary with respect to the forms of their referents.  *By definition*, 
>the symbols in the Nonsense room have no referents, so obviously the room, 
>person, etc. cannot know the referents of them.  In both cases there is no 
>"understanding" by anyone or thing.  If you claim there is for the Chinese
>Room, then it has to be coming from the Chinese characters themselves or the
>particular set of rules.  But in what way (besides their literal forms) 
>are these different than the nonsense characters and rules?  If you think
>there is some difference other than literal form, then where is it?

The important difference is that the Chinese Room System is capable of
performing the reasoning, remembering, and learning necessary to
simulate human intelligence and the Nonsense Room System is not.  The
foreign language aspect of the Chinese Room scenario is really a red
herring -- the important thing is that the room's replies simulate
those of an intelligent being.

For example, compare the following two dialogues:

(In Chinese)

Human: My name is John.
CRS: Hello, John.
Human: What is 2+2?
CRS: That's easy, 4.
Human: What's the next number in this sequence -- 2, 5, 10, 17?
CRS: Hmmm.  Hold on a second.  <pause>  That looks like x^2+1, so I'll
     guess 26.  Am I right?
Human: You're right.
CRS: Great.
Human: What's my name?
CRS: You said it was John.
Human: What's my gender?
CRS: Well, John is usually a man's name, so you're probably male.

(In Nonsense)

Human: My name is John.
NRS: Gsyn xbso slnf
Human: What is 2+2?
NRS: Snorfgag dnsnlio
Human: What's the next number in this sequence -- 2, 5, 10, 17?
NRS: Blapksnc idoslinnbc oksd
Human: What's my name?
NRS: Thdooosn ndallolppll
Human: What's my gender?
NRS: Cosssshbnfn nnkkil kolpoienn

Now, either you can translate the responses of the Nonsense Room
System into coherent and relevant replies or you can't.  If you can,
then it's no longer speaking nonsense -- it's speaking the new
language you've just invented.  If you can't, then the difference
between it and the Chinese Room is clear.


