Information Technology Reference
In-Depth Information
It is perfectly legitimate to use computational methods for some first and prelim-
inary evaluations, as we use the thermometer, the speedometer, the yardstick. When
a distance is measured as five meters, some of us say, “oh, I can long-jump this
easily”. Others will never make it. But all try very hard.
When the temperature in a room is measured as 22 degrees Celsius, some re-
act with “too hot for me”, others with “rather cool after a while”. Measure, value;
evaluation, judgement.
And let us not forget, how you, Harold, continue after your optimistic remark
about what the machine might be capable of. You say that you would still take a
look before, upon the program's evaluation, you delete the file. . .
PG:
I think that this is the kind of discussion that can always be paused but never
ended. For now I'd be happy just to clarify what the differences are.
If it turns out that non-trivial computational aesthetic evaluation is impossible,
that in itself would be worth better understanding. It seems to me such a statement
might come in two forms. There might be some kind of formal sense, or there might
be an engineering analysis leading to absurdly expensive, or quantitatively impossi-
ble, practical requirements.
Frieder seems to lean towards the former by saying that aesthetic evaluation
would have to be formally computable, but is not. But this leads to (in my mind)
an even more interesting question. How is it that the mind is capable of “comput-
ing” the uncomputable? Is the mind more than the result of a mechanistic brain?
And if the objection is more practical and in the realm of engineering a similar
question is raised. What aspect of the mechanistic brain can we know to be beyond
the reach of human engineering? How is it that nature has brought the costs and
quantities within reach in a way we will never be able to duplicate?
The strongest objection, to me, would also be the one that claims the least, i.e. that
computational evaluation as an engineering challenge is impossible
for the time be-
ing.
Maybe it will be within reach in ...10 years? 50 years? 100 years?
But if the operative objection is this last one it changes the entire conversation.
Because then computational aesthetic evaluation is possible in principle and merely
contingent. All discussions of creativity should allow for it in principle.
Frieder also mentions that, “Judgement is different from evaluation”. In our
Dagstuhl discussion Margaret Boden rejected such a notion out of hand. Perhaps
they are referring to two different kinds of judgement, or two different kinds of
evaluation, or both. In any case this confirms in my mind that the language involved
will need more precision than everyday speech, and technical definitions are prob-
ably called for. For example, when a human takes a given work of art and merely
classifies it to an art movement, can that be called “evaluation” or should some other
word be used?
Finally there is a bit of a paradox worth pointing out here. Most attempts to define
creativity I heard at the Dagstuhl workshop included a provision that the innovation
must not only be new but it must also be of value. Now if computational aesthetic
evaluation is more or less impossible does this mean computational creativity is
impossible? Or does this mean a computer can be creative without being able to
measure the value of its own output?
