From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!yale.edu!jvnc.net!nuscc!maclane!smoliar Tue Jan 21 09:26:31 EST 1992
Article 2815 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!yale.edu!jvnc.net!nuscc!maclane!smoliar
>From: smoliar@maclane.iss.nus.sg (stephen smoliar)
Subject: Re: Searle Agrees with Strong AI?
Message-ID: <1992Jan17.013013.8681@nuscc.nus.sg>
Summary: from artificial intelligence to artificial life?
Sender: usenet@nuscc.nus.sg
Reply-To: Smoliar@ISS.nus.sg (Stephen Smoliar)
Organization: Institute of Systems Science, NUS, Singapore
References: <1992Jan16.054716.14332@oracorp.com> <1992Jan16.145637.26097@news.media.mit.edu>
Date: Fri, 17 Jan 1992 01:30:13 GMT

Every time I start feeling saturated with this bulletin board and entertain
serious notions of giving it up, Marvin comes along and starts to make life
interesting again.  In article <1992Jan16.145637.26097@news.media.mit.edu>
minsky@media.mit.edu (Marvin Minsky) writes:
>  How about this: let's let Searle
>off the hook for a moment, be asking this question:
>
>	If we could build a machine that is suitably reactive, and can
>	assemble raw materials so as to make working copies of itself
>	would the resulting machine be ALIVE?
>
>In  other words, is "understanding" analogous to "living" in the old
>vitalist controversies?
>
This COULD be interpreted as a proposal to start a comp.AL.philosophy bulletin
board;  but I suspect we can accommodate "artificial life" here, at least for
a while!  I am particularly interested in this question because I have been
hard at work on a review of the two volumes of artificial life proceedings
published by the Santa Fe Institute (along with two hours of video supplement).
Now that my thoughts are beginning to come into shape, I need to bounce them
around a bit.  (Thanks, Marvin.)

Before getting to Marvin's specific question about building a machine, I want
to make the observation that there seems to be a lot of work in graphics which
is being passed off as artificial life.  The late Aristid Lindenmayer,
Przemyslaw Prusinkiewicz, and their colleagues were represented at both
New Mexico workshops for the work they did on using L-systems to develop
graphic images of various organic cellular structures.  In the first
proceedings volume their work was represented as a contribution to
embryology;  and in the second it was classified as "development."
Having now read these two papers for myself, I would say they are
nothing of the sort.  They tell you some very neat things about how
you can come up with powerful life-like graphics;  but they offer no
evidence that a better understanding of these graphics will lead to
a better understanding of either embryology or post-natal physical
development.  (For that matter, they do not deal with the flip side
of the coin, either:  whether or not a better understanding of the
underlying biology would contribute to being able to construct better
graphics.)

There is an important lesson to be learned from these two papers;  and it is
that it is very easy to confuse the APPEARANCE of life with life, itself.  This
should immediately lead us to ask if the really IS more to life than
appearances and whether or not we can characterize what that "more" is.
The vitalists take the easy way out.  They can always consult Genesis for
an account of those vegetables and animals which were endowed with life and
write the whole thing off to Divine Decision.  In posing his question
informally, Marvin has stepped into the most viable alternative to vitalism
I have found so far--the suggestion that life may be a characteristic of
certain behavioral FUNCTIONS, including reacting to the surrounding environment
and reproducing itself.  This seems like a good analogy to Turing's position.
Rather than retreat into a philosophical quagmire over whether a machine can
REALLY think, set up a far simpler scenario, that of the Imitation Game, and
ask if a machine could be dropped into it in a manner which would be
"convincing" to a sufficient number of observers.  Some members of the
artificial life camp have taken this approach.  Dave Jefferson, for example,
has worked on enumerating a suitable list of functions and then using that list
as a standard against which to evaluate his own work.

Having said all that, however, it is worth asking whether or not we have really
gotten beyond APPEARANCES.  After all, if we are talking about observing how a
machine behaves, are we not more concerned with what it looks like it is doing,
rather than what it is ACTUALLY doing?  Perhaps this really is the case, which
means that I may be too harsh in my assessment of the Lindenmayer group.

However, there is still a step which we can take beyond appearances;  and that
is the matter of MODELING, which was not really addressed by Turing and is not
part of the question which Marvin formulated above.  My primary attack on using
L-systems to make pretty graphics is that it does not help us to answer
questions about what is really going on during the pre-natal and post-natal
development of various organisms.  That is what modeling is all about--coming
up with artifacts which help us answer such questions.  If I am confronted with
a physical machine which reproduces itself (or, to return to a favorite Searle
hobby-horse, which is capable of acquiring, consuming, and processing energy
resources in a manner which gives the appearance of feeding, digesting, and
eliminating), I am less interested in whether or not I, or anyone else, is
willing to SAY that it is alive and more interested in whether or not it can
answer any particular questions I may have about the reproductive or metabolic
processes of any "real" organisms which happen to interest me.

At this point I suspect we cannot avoid intrusion from the philosophers again.
If it is a machine, society is not going to get upset if I stick all sorts of
laboratory sensors and effectors into it or turn it off to take it apart
without absolute confidence that I can put it back together and turn it
on again.  We do things like that to machines, and society accepts this
as part of proper behavior.  However, if society now judges the sort of
artifact Marvin has hypothesized to be ALIVE, the rules of the game suddenly
change.  Suddenly I may find my laboratory being picketed by an army of animal
rights activists;  and, from their point of view, they may well be justified
in being outraged by my behavior.  After all, if I turn it off and mess things
up to the point where I cannot turn it back on again, have I killed it?

Some of these philosophical questions came up during the second artificial life
workshop in Santa Fe.  My personal feeling is that the discussion of these
questions has been poorly represented in the contents of the ARTIFICIAL LIFE
II proceedings volume.  However, I am not sure that any firm conclusions were
drawn which would justify scientific publication.  All that really emerged was
a general consensus that people who do this sort of work should proceed with
caution.  Needless to say, no one proposed a behavioral code for what it would
MEAN to "proceed with caution;"  and I am not sure that I have enough
confidence in any of the current researchers to accept any such code
they might propose.  Perhaps we should not turn to scientists for such
matters and should, instead, return to those who wrote fiction before
science had made their visions realizable--writers such as Mary Shelley
and Isaac Asimov.
-- 
Stephen W. Smoliar; Institute of Systems Science
National University of Singapore; Heng Mui Keng Terrace
Kent Ridge, SINGAPORE 0511
Internet:  smoliar@iss.nus.sg


