Information Technology Reference
In-Depth Information
As discussed elsewhere in this topic (e.g. Chap. 5), becoming an expert or vir-
tuoso in a particular medium normally takes many years of intense practice and
immersion. As expertise and virtuosity mature, so does evaluation: the two appear
to go hand-in-hand. Knowledge and experience emerge as decisive factors in pro-
ducing artefacts of high creative value.
With these statements and questions forming a background, let us now proceed
to the discussion.
4.3 A Conversation on Evaluation
The participants are (in order of appearance and identified by their initials), Harold
Cohen (HC), Frieder Nake (FN), David Brown (DB), Jon McCormack (JM), Paul
Brown (PB) and Philip Galanter (PG).
The conversation begins with a discussion about the aesthetic evaluation of art
by people and computers
.
Harold Cohen (HC):
I sometimes wonder whether Western culture hasn't gen-
erated more art evaluation than art over the past few hundred years. How much of it
is known outside the art world is another matter. It is worthwhile to make clear that
aesthetic evaluation has little to do with conformance to the set of “rules” still being
widely taught in art colleges.
As to the evaluation of aesthetics computationally, I confess to paying little at-
tention to what's going on outside my studio, but I'd be very surprised to learn that
there's a rich enough history of practical work to fill a topic. Why is there so little
history? To begin with, AI is still not at a stage where a program could accumulate
enough relevant knowledge about an object it didn't make itself to make a non-
trivial evaluation, so the discourse is limited, necessarily, to art-making programs,
of which there have been relatively few. (I'm unclear about whether the same limi-
tation would apply in other forms: music, for example.)
All of my own sporadic forays until now have been non-starters. But once I relin-
quish the notion of program autonomy and accept that the program is working with
and for me, it becomes clear that it is capable of exercising (my) aesthetic judge-
ment. And it does, to a point. But it's exercised on the work-in-progress, not on the
finished work. Thus, it doesn't wait to tell me that an image has too much grey;
it evaluates and corrects as it proceeds, provided that I can tell it how much grey
is enough. That's a trivial example; one step up I'd need to say how much grey is
enough relative to something else. Even if I could find a way of identifying amount-
of-grey as an evaluation issue, and say what to do about such issues generally, there
is still the problem that they are a moving target. That target moves in the human
domain.
Unfortunately, it's a lot easier to say what's wrong with an image than to say what
makes it special. I'm looking for the images that are
transcendent;
which means, by
definition, that I don't know what it is that makes them special and don't know how
