Build a database, randomly assemble.

Recurse.

Use whatever language generator deemed appropriate* to assign proper titles for the newly-evolved art movement.

*this may be delegated to a meat-computer, in pity.

Before that becomes the main problem, that other problem must be solved where you give it your picture and the Mona Lisa whose style is to be copied and the way it copies the style is it paints her eyebrow atop your mouth. 'Cause it's not like this thing can really distinguish between "style" and "content". What it does is it pastes patches from the "style" image onto the "content" image. Sometimes the result is interesting, much of the time it's between boring and terrifying. It's cool and all, but it's not much more intelligent than your favorite Photoshop filter.

"Going the way of the dodo..."