From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!wupost!uunet!tdatirv!sarima Mon Mar  9 18:34:44 EST 1992
Article 4214 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!wupost!uunet!tdatirv!sarima
>From: sarima@tdatirv.UUCP (Stanley Friesen)
Newsgroups: comp.ai.philosophy
Subject: Re: Definition of understanding
Message-ID: <464@tdatirv.UUCP>
Date: 2 Mar 92 22:07:01 GMT
References: <1992Feb22.181122.12088@oracorp.com> <6254@skye.ed.ac.uk> <1992Feb26.102122.22893@nuscc.nus.sg> <6303@skye.ed.ac.uk>
Reply-To: sarima@tdatirv.UUCP (Stanley Friesen)
Organization: Teradata Corp., Irvine
Lines: 42

In article <6303@skye.ed.ac.uk> jeff@aiai.ed.ac.uk (Jeff Dalton) writes:
|
|I suppose it's possible to quibble endlessly about the word
|"understand" and thus avoid ever considering Searle's argument.
|But all that's actually required to make the first step is
|that you can distinguish between a case where you understand
|a language and one where you don't.  A full understanding
|of everything that understanding involves is not required,
|nor is a an understanding of all sorts of other senses of
|the word "understand".

I beg to disagree here.

I can tell (usually) when *I* understand something, however when it comes
to telling if *you* understand something I must make an educated guess
based on your responses.

Now Searle has introduced this system called the Chinese Room.  How do I
tell whether it, as a whole, cna understand?  I am not it, so I do *not*
know what sorts of subjective feelings it may or may not have.  What feelings
its hardware (the man) might have are *irrelevant* except to the extent they
instantiate feelings in the whole system.

So, how do I tell if the CR has understanding?  Well, you want to rule out
using the responses (aka The Turing Test).  Fine, now we have *no* way of
telling until such time as we have a clear specification of just what it
means to understand something.


Thus the meaning and sense of the word 'understand' are *critical* to the
answer.   In fact that is one of two critical pieces of information that
Searle has left out.  The other is a general idea of what sort of
implementation the CR uses.

Once we have *these* *two* *things* we can simply compare the implementation
to the specification to see if they match.  If they do the CR understands,
if they do not, it does not.  Quite simple?  Not really, both are beyond
our current level of technical expertise.
-- 
---------------
uunet!tdatirv!sarima				(Stanley Friesen)



