Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!yale!quasi-eli!cs.yale.edu!mcdermott-drew
From: mcdermott-drew@cs.yale.edu (Drew McDermott)
Newsgroups: comp.ai.philosophy
Subject: Re: Testing for machine consciousness
Summary: It's straightforward
Message-ID: <27037@cs.yale.edu>
Date: 2 Nov 90 21:36:06 GMT
References: <1990Oct31.023922.13795@watdragon.waterloo.edu> <1990Oct31.142817.1999@canon.co.uk> <1990Nov1.091704.6831@canon.co.uk>
Sender: news@cs.yale.edu
Organization: Yale University Computer Science Dept., New Haven, CT 06520-2158
Lines: 72
Nntp-Posting-Host: aden.ai.cs.yale.edu
Originator: dvm@aden.CS.Yale.Edu

Quoting Robin Faichney <rjf@canon.co.uk>:

   >We experience things, and are 'programmed' to identify with other
   >humans, ie to believe them essentially identical to ourselves, so we
   >believe that they experience things.  Some things we do not identify
   >with, ie we do not believe they experience anything.  

Surely there can be no question whether humans actually experience
things.  If you ask me to doubt whether I actually experience
anything, I will experience several things, including puzzlement!

Your paragraph is a puzzle already.  The first three words grant that
"We experience things."  But you say "we ... believe that *they*
experience things."  (italics added)  Surely we=they here?

   >... The main consequence for AI is probably that, if you disregard the
   >inherited predisposition to attribute consciousness to (ie identify
   >with) other humans, such attribution is at best arbitrary and at worst
   >meaningless.  Unless there is something seriously wrong with my
   >account, there can never be a good reason to seriously attribute
   >consciousness to a machine.

There is something seriously wrong; see above.

   >As Eliza demonstrated, people will quite readily interact with a
   >machine as if there was 'a real person in there', even when they know
   >there is not.  

I am completely confident that Eliza experiences nothing, regardless
of how comforting it can be to talk to it.

    I think that this phenomenon is interestingly similar to
   >the kind of suspension of disbelief which occurs when we are 'taken in'
   >by a good film, play, book, etc.  We know that the characters are not
   >real, but can feel, to some extent, as if they were.  I'd put quite a
   >lot of money on the proposition that this will be the main way people
   >will interact with computers ....

Interesting observation; no doubt correct.

But I still maintain that the fuss about "testing for consciousness"
is misguided.  Here's how it will work: We will figure out (via
modeling and vivisection) what's going on in people that counts as
consciousness.  We will then duplicate that something artificially and
verify that the resulting system is also conscious.  (It will have
strong opinions about the way it works that are isomorphic to ours
about ourselves).  We will also no doubt produce so many variations on
the theme that our concept of mind will change considerably by the
time we're done.  (As Chris Malcolm has emphasized in this thread.)  No
particular definition of consciousness will emerge, and the desire for
one will evaporate.  What will emerge is a good understanding of how
to manipulate different aspects of consciousness.  So, if you want a
robot with, say, qualia but no free will, you can have it.

Let me hasten to add that this scenario is not inevitable.  It *could*
turn out that a much more radical revision of our conceptual framework
results from our investigations, so that, e.g., we end up saying that
no system ever actually experiences anything.  Or it could turn out
that we never get a satisfactory theory of consciousness, and it
remains a mystery.  But if the theory of consciousness evolves as
previous scientific theories have evolved, then, I claim, we need have
no qualms about any special methodological problems with it.

   >I'd also like to humbly suggest that future generations of AI workers
   >will look back with amusement and bewilderment at such arguments as to
   >whether a machine could be conscious, much as we do at the medieval
   >arguments about the number of angels which could dance on the head of a
   >pin.

I agree, but for somewhat different reasons.

                                             -- Drew McDermott