Path: utzoo!utgpu!news-server.csri.toronto.edu!csri.toronto.edu!me
Newsgroups: comp.ai.philosophy
From: me@csri.toronto.edu (Daniel R. Simon)
Subject: Re: Testing for machine consciousness (was Re: emergent properties)
Message-ID: <1990Oct11.161350.16127@jarvis.csri.toronto.edu>
Organization: CSRI, University of Toronto
References: <1990Sep29.213139.2876@watdragon.waterloo.edu> <3499@media-lab.MEDIA.MIT.EDU> <1990Oct4.154655.23004@canon.co.uk> <7@tdatirv.UUCP> <1990Oct8.120927.8648@canon.co.uk>
Date: 11 Oct 90 20:13:50 GMT
Lines: 70

Perhaps some recent experiences of mine will add to the current debates over 
consciousness and emergence.

I have just returned from a visit with the graphics group at a rather
prestigious academic institution which shall remain nameless.  I say "graphics
group", but I should note that they have changed their name to "Laboratory for
Artificial Appearance", and consider themselves to be grappling with much 
deeper problems than the mere rendering of amusing images (although the bulk
of their lucrative research work is in practice of precisely this variety).

Their transformation from "graphics group" into LAA began with an article 
published some years ago in an obscure graphics journal, but now considered a
classic in the field.  It proposed the following thought experiment:  suppose
that a television camera were pointed at a person sitting against a blank 
background; the resulting image (which might be either black and white or 
colour, and either still or moving, as the experimenter prefers) is displayed 
on an ordinary CRT screen in a different room.  Adjacent to this screen sits 
another screen displaying an artificially rendered image (again, either black
and white or colour, and either still or moving, to match the image on the 
first screen).  Individuals are invited into the room to examine the screens,
and to try to determine which screen is displaying the image of a real person.
The article provocatively asks whether, if even very sharp-eyed viewers are 
unable to spot the signs of computer rendering in the artificial image, it can
credibly be denied that the artificial image is in fact, the image, or (in some
strict sense) the "appearance" of a human being.

The consequences for the field of this playfully-named "blurring test" have, it
seems, been profound.  Numerous extensions have been proposed to the original
test, including the possibility for aural or tactile components to the images, 
and perhaps even interaction.  Graphics researchers who still have difficulty 
producing realistic images of trees have found ample backing for their efforts
to theorize about how to produce flawless renderings of the human form.  
Philosophical debates have abounded concerning the definition of appearance,
appropriate goals for artificial appearance researchers, and most of all, what
properties artificial appearance shares with its "natural" counterpart.

For example, during my stay at the LAA, a heated discussion ensued concerning
the property of beauty, and in particular over whether artificially-rendered
humans could be considered beautiful.  Many, of course, argued that no mere
collection of pixels could ever be considered beautiful in the sense that real,
live humans could be; much was made of the significant role of context in 
beauty.  Some suggested that the sceptics were defining beauty too narrowly;
was there no beauty, they asked, in the Mona Lisa or the Venus de Milo, not to
mention Mondrian's "Broadway Boogie-Woogie"?  The middle position, I gathered, 
was that, at the very least, an image which passed the blurring test deserved 
to be considered as beautiful as if it were a real person, although it was 
widely conceded that no artificial image yet produced was worthy of the 
attribution.

Two aspects of this discussion make it, I think, relevant to the current ones 
in this newsgroup:  firstly, the sceptical "hard line" was frequently rebutted 
with references to the idea of beauty being an "emergent" property.  Hence, it 
was argued, a collection of pixels may well be "humanly" beautiful, although
it was composed of individual unbeautiful parts.  Secondly, their use of the
blurring test was, for me, highly evocative of the current debate; for example,
they frequently asked how, if one could deny the beauty of a computer-generated
image indistinguishable in "appearance" from a really beautiful woman, one 
could still say with even reasonable certainly that a beautiful woman herself
is really beautiful; could not her appearance, too, be secretly composed of 
minute dots of colour?

My only contribution to the debate was to remark once, half in jest, that I
had always believed beauty to be in the eye of the beholder.  There was an
awkward silence, a few people coughed embarrassedly, and after a few moments
the conversation continued as before.  Until I left, those were the last words
I dared speak on the subject.


"There *is* confusion worse than death"		Daniel R. Simon
			     -Tennyson		(me@theory.toronto.edu)