Decoupling Art and Affluence
Harold Cohen, University of California San Diego
2001 International Joint Conference on Artificial Intelligence
Kurzweil CyberArt - demo applet, screen shots
Article on AARON at ViewingSpace.com
ETA, 00:15 CDT Thu 05 May 2005: jereeeza makes a valid critique:
Ah, AARON. What does it obsess about? What bothers it? What makes it happy? What metaphorical intent drives it towards pictures of people and plants? Whence the symbolism of those particular colours?
The English language's painful lack of distinction between art and exercise strikes again.
I have no definite answers to these questions, of course. I am also not the person who claims it is producing art (as opposed to a program that generates original drawings), though I do cite the title and billing. It is putatively referred to as a system for automated generation of art, with terms such as "creative" and "original" being bandied about loosely by non-artists and nontechnical viewers of its output.
As for obsessions, disturbances, happiness: you artists know better than I do how these are or are not necessary conditions for a drawing to constitute art. This version1 of AARON is programmed quite literally to generate drawings that are representational in that it has specific routines for human figures, potted plants, etc. I would assert that the aspect of its "metaphorical intent" that "motivates" it to include a closed curve as interesting, representational, or part of a figure belongs entirely to its programmer. In this version, at least: there are generative systems that are motivated by fitness judgements, both human, mathematically defined, even constrained and semirandom.
Are these "real" aesthetics? You are, of course, entitled to assert that they are not. I think it depends, but as Cohen himself noted, he does not consider AARON to have a level of creativity that compares to its programmer; specifically, it has no learning, adaptive, or self-modifying aspect. More to the point, it is completely open-loop: it does not interact with its environment at all; it cannot "see" what it has drawn in that aside from coloring its forms, it does not perceive or evaluate the representation, nor does it modify either the image or its own internal representation on that basis.
1 And here's my pet peeve: I think he should note which version is which, numerically or otherwise, whether he distributes its source or no.