I’ve been experimenting with image-generating neural networks, which look at a
bunch of images and via trial and error gradually learn to produce more like
them. Or at least “like them” according to the neural net’s own interpretation
of realistic, which is usually missing a lot.
Last time
