Information Technology Reference
In-Depth Information
A second program took this statistical description of an image (really: an infinity
of images) and distributed colours into a quadtree structure such that the prescribed
(just calculated) frequencies of colours were obeyed. I called the quadtree structure
“the topology of the image”.
I guess it was one of the most powerful programs ever in computer art, and cer-
tainly of its early phase. The program showed how little you achieve this way. As
Harold says, you can use such dynamic evaluative measures during the generative
process. That's all. Anything beyond this is human value judgement.
Phil Galanter has shared some of the scepticism of others, but says giant leaps
are not to be expected. But baby-steps should be tried just to see where they get the
baby. Yes, dear Phil, what is there left to do other than doing baby steps. So let us
get into those pink knitted tiny shoes that mothers like to put their baby's feet into
and move on from there.
David Brown (DB):
I think that an analysis of existing methods in order to in-
fluence the output of computational systems—via some embedded knowledge (such
as rules)—
is
a useful thing to do.
My experience in the design world suggests that you'll find a lot of people who
had “techniques looking for a problem”—i.e. the method of evaluation is shaped by
their tool.
I think it is better to analyse the problem and then look for techniques. For ex-
ample, what kinds of evaluations affecting creativity are made during synthesis and
what kinds of techniques can make these evaluations? Additionally, what kinds of
evaluations can be applied to the descriptions of resulting artefacts, always assuming
that all necessary sensing is in place.
For creative evaluation, newness and surprise are key to people judging some-
thing as being creative. But judging both of these computationally is tricky, espe-
cially during synthesis.
Focusing on learning is putting the cart before the horse. Focusing on a belief
that something is “impossible” is not letting either out of the stable: a great way to
reduce discovery of, and understanding about, the ingredients that lead to creative
artefacts. By taking each challenge and looking at how it might be tackled we can
make systematic progress.
Can we get a system to figure out that a blue widget isn't much different from a
green widget, even if in some sense it is “new”? How can different types of newness
be evaluated? Can a system predict how much a “newer” choice during synthesis
will affect the judgement of the creativity of the finished product?
We take questions such as these and look for techniques that might help. For ex-
ample, could we use the web for assessing newness? Could we take a representation
of an artefact that has structural, behavioural and functional components and use
that to decide a degrees of newness? Could fuzzy matching techniques be used to
detect similarity and therefore newness? And so on. . .
Jon McCormack (JM):
This discussion has made a number of claims as to why
objective aesthetic measures seem impossible for an individual or machine. Never-
theless, I do think there is some basis for looking at aesthetic commonality particular
