From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!mp.cs.niu.edu!ux1.cso.uiuc.edu!rbrown Wed Apr 22 12:04:18 EDT 1992
Article 5171 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!mp.cs.niu.edu!ux1.cso.uiuc.edu!rbrown
>From: rbrown@ncsa.uiuc.edu (Rich Brown -- Neither)
Newsgroups: comp.ai.philosophy
Subject: Re: Intelligence, awareness, and esthetics
Summary: "I don't know much about information theory, but I know what I like!"
Message-ID: <1992Apr21.155531.23910@ncsa.uiuc.edu>
Date: 21 Apr 92 15:55:17 GMT
Sender: rbrown@ncsa.uiuc.edu (Rich Brown Cray Operations)
Organization: Nat'l Ctr for Supercomp App (NCSA) @ University of Illinois
Lines: 88
Originator: rbrown@mars.ncsa.uiuc.edu

Shannon's _Information_Theory_and_Aesthetic_Perception_ contains some
interesting analyses of how one might actually digitize the rate of information
transfer associated with a work of art.  It has to do with viewing as a single
system the object, the senses of the person apprehending the object, and the
information conveyed.  More precisely, he is concerned with the _rate_ at which
information is conveyed, rather than the _amount_ of information being
conveyed.  In information-thoretic terms it is a model something like:
	Transmitter-->Information Channel-->Receiver.

So, for example, when a symphony is performed, the model includes the physical
limits of the human ear, the acoustics of the hall, the pitch, volume, timbre
and clarity of the tones produced by each instrument, etc.  When the same
symphony was performed by different orchestras under different conductors, and
digitized, Shannon had a complicated method of deciding at what _rate_ a single
bit of information was transferred to the human receiver.  He stressed that it
had little to do with the tempo at which the piece was played, nor the sheer
number of tones heard at any one moment, nor the volume, nor pitch
differentials (all these things certainly produce _more_ information) but
rather was a measure (supposedly) of the _rate_ at which bits flowed through
the system, using these various things, as well as the limits of the human ear,
as constraints with respect to the orchestra-->hall-->listener system.

Shannon claimed that the symphonies as played by the "greats" produced higher
numbers than the same symphonies conducted/performed by less skilled musicians.

With respect to painting, eye movements were tracked as viewers gazed upon
various works of art, and for these too Shannon claimed that his formula, when
applied to the works of the great masters, consistantly produced higher numbers
than what most folks considered poorer art.  Here he used constraints such as:
Smallest amount of light detectable by the human eye, largest amount (the
saturation point -- the amount of light beyond which the eye could not detect
any further brightness) and the smallest incremental difference in brightness
which can be detected by the eye.  Similarly, constraints were used for the
eye's ability to detect differences in hue, and texture, and the size of
regions in the painting (i.e, what pixel size is indistinguishable?).  The
claim was that the _rate_ at which information is transferred along the channel
to the receiver is higher for works by the great artists (and by a few
institutionalized schizophrenics!) than it is for, say, black velvet Elvi.

The implication was that it is completely possible to objectify aesthetics, and
so justify a claim that some work of art was "better" than another...

Of course, everyone knows that art is subjective, don't they?  Or maybe some
people just like poor art better than good art.  :-)

I've deliberately avoided giving MHO here, 'cause I don't want to get my ass
flamed, but I will be very entertained if some of you have read or will read
this book (sorry -- I think it was published here at the U of I, but I could be
mistaken) and flame each other heavily about _your_ opinions of it.  I feel
fairly certain that some readers will react violently against such a theory,
while others may discover that it seems entirely plausible.

If scientists fall into two camps over this issue, does that mean that science
is subjective as well?  "I don't know much about information theory, but I know
what I like!"

References follow:

rickert@mp.cs.niu.edu (Neil Rickert) writes:

>In article <1992Apr20.191345.27706@javelin.sim.es.com> biesel@javelin.sim.es.com (Heiner Biesel) writes:
>>This suggests an alternative to the Turing test: a computer can be
>>said to exhibit human intelligence and awareness if it is capable of
>>producing a work of art which finds both wide acceptance among art
>>lovers, and is indistinguishable from similar pieces or art produced
>>by human artists.

> However, please keep in mind that many works of art produced by humans in
>the 20th century would have been met with utter revulsion from the art lovers
>of the 19th century.  What is considered good art is highly sensitive to
>cultural influences.  This does not necessarily negate your idea - except
>you are now testing cultural adaptation rather than intelligence.

True, much of what moves us in great art is cultural and learned. Fortunately,
we have centuries of styles, and many cultures to draw upon. The essence of
the proposal is to base a test upon an esthetic criterion, rather than a
primarily linguistic one. It is not by any means perfect, but for some lucky
few unequivocal, to my mind an essential aspect of any such test.

Regards,
       Heiner biesel@thrall.sim.es.com


-- 
+++++++++++++++++++++++++++++++++++++++++++++++
Rich Brown, UNICOS System Administrator
National Center for Supercomputing Applications
University of Illinois at Urbana-Champaign


