Kandinsky's art has never really been my cup of tea. I dig abstract art, but I need a little more prettiness in my abstractness. Kandinsky was more into the sort of deeply angular weirdness that inspired many devoted postmodern followers, and probably a few Yes album covers. (Interestingly, Kandinsky was something of a synesthesiac, and claimed that he saw color when he listened to music.)

Anyway, Robert Peake is both a very cool programmer and a fan of Kandinsky, so he created a virtual-artist artificial-intelligence program that produces Kandinsky-like drawings -- including the one above! Just pump in a few variables -- how many shapes you want, the size and width of the picture -- and it'll kick out insto-art. His FAQ is a blast to read, and includes this Q&A:

C'mon, isn't this really just a bunch of random shapes?

Well, yes and no. I have already begun to make aesthetic decisions, like assigning a color palette to the shading and constraining how far the shapes can be flung. Also, the "randomSymphonic" class does something very common in Kandinsky's work - arranges either 3, 5 or 7 shapes in a vertical pattern. Usually, these look like brush strokes. It's a start.

So, already the beginnings of human aesthetics are poking their way in to the picture. I'll need to bear these in mind as a baseline when I start applying more advanced principles. The ultimate goal of this project is to apply and then test (via human feedback) a wide variety of aesthetic principles. Because it is web-native, the potential for getting a large sample size for human data on what "is and is not art" as well as mapping this data to precise heuristics makes this project interesting.

(Thanks to Slashdot for this one!)