Style Transfer In Reverse

How AI helps unlock the secrets of Old Master and modernist paintings
Jan 2020, Ars Technica

Agog, askance, agog; I am still unable to look at a series of style-transfered images without unfettered vacillation. How is this real? This can't be real. But it's not that hard; if you just look at the way the...no. This can't be real.

I went to college for art history, so I spent 4 years looking for meaning in visual patterns. Never did I think computers would be able to do this. It's one of those things that will be human all the way til the end. Wrong. Machine learning, of the Deep kind, using neural networks, is not like your grandmother's computer. The wet parts, the sloppy parts, the imperfect, uncountable, and very human parts are now under the cold hard gaze of Technology's highest achievement, artificial intelligence.

A properly trained program can take a picture of Sesame Street's Big Bird, run it through a hundred Francis Bacon paintings, and return an avian nightmare phantasm. It can even abstract a tree in the style of peri-abstraction-Mondrian!

And as one would expect, it can also look at a hundred paintings and tell you if they're really from the same artist. One step further and it can actually fill-in missing pieces of damaged work.

So it sounds like computers can now make the art themselves. Finally, now the humans can move on to more important tasks.

Notes:
Raiders of the Lost Art.
Anthony Bourached, George Cann. arXiv, 10 Sep 2019. https://arxiv.org/abs/1909.05677

Semi Automatic Artifact Generator
Oct 2018, Network Address

Initial musings on genenrative adversarial art etc.