Train a GAN with images with intent, then have the computer produce an image that lacks intent of an artist but imitates the visual amenity of the input works? Art-in-the-shell??