Does the world need a Hilla and Bernd Becher embed? Who knows. But it has one now. Trained on 36 black and white photographs, it does a reasonably capable job of generating structures or giving structural qualities to an image.