In Artificial Intelligence, we try to automate analytic cognitive tasks, such as problem solving, image interpretation, and language understanding. Artificial Creativity may be viewed as the counterpart of A.I. which is concerned with generative tasks, such as artificial art, story generation, and playing non-competitive games.
The success criterion for Artificial Creativity is difficult to pinpoint. For the tasks it addresses, we cannot explicitly specify when a solution would be “correct” rather than “incorrect. Instead, we are interested in the variety and unpredictability of the set of system outputs.
Focussing on automatic image generation (“artificial art”), this talk will explore different techniques for characterizing image classes, and strategies for achieving variety and unpredictability. Two major approaches will be compared: (1) constructing generative grammars or image algebras, and (2) exploiting the phenomenon of “emergent form” as it is often encountered in Artificial Life.