From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!wupost!uunet!psinntp!scylla!daryl Fri Jan 31 10:27:22 EST 1992
Article 3303 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!wupost!uunet!psinntp!scylla!daryl
>From: daryl@oracorp.com
Subject: Re: Evidence that would falsify strong AI.
Message-ID: <1992Jan30.172057.7114@oracorp.com>
Organization: ORA Corporation
Date: Thu, 30 Jan 1992 17:20:57 GMT

Jeff Dalton writes: (Responding to Colin Garety)

>>> What I find strange in all this is that anyone should find it
>>> a mystery what "understand Chinese" means.  Can these people
>>> really no distinguish between such things as reading a book
>>> written in a language they know and reading one in a language
>>> they don't? Do they really think they can't answer this
>>> question until someone tells them what "know" means?
>>
>>I have little trouble figuring out the extent to which I understand
>>Chinese. The chinese room leads to the question "How can I tell whether
>>you understand Chinese?"

>The question is: do you have some difficulty in knowing what
>understand means?

I believe that he already answered that. He knows what it means for
Colin Garety to understand something, but he doesn't know what it
means for you to understand something. It is quite clear from the
endless debates on this subject that there is no agreed-upon meaning
of what it means for someone to understand.

You act as if your personal experience of understanding is sufficient
to be able to successfully apply the word to others. I disagree. My
personal, subjective feeling of understanding is not a two-place
predicate--understands(X,Y)--it is a one-place predicate:
I-understand(Y). The generalization from a one-place predicate that I
experience directly to a two-place predicate that I can have no direct
experience of is extremely difficult for me. Frankly, I don't have any
idea how to do it, short of using some functional or behavioral
meaning to "understand", which definitely does not capture any of the
subjective feeling of understanding.

> But some people will call any definition ill-defined or unclear
> unless it directly corresponds to some test.

I would say that a definition of understanding is unclear if, given a
complete description of a system, it is not possible for you to
explain why the system does or does not meet the definition. We have
bandied about the prime example: a gigantic lookup table program. I
think that by now it should be pretty clear how this thing is supposed
to work. What is the argument that it doesn't meet your definition of
"understanding"? David Chalmers claims that such an argument has been
published, but I would like to know how you would argue it.

Daryl McCullough
ORA Corp.
Ithaca, NY





