You ever wondered if we could take a neural net, let it purposely hit the edge of its training data, and use that intentional corruption as a new aesthetic language?