AI in the News — #18

Robot Forgery (NPR)

An AI has produced a very plausible fake Rembrandt. In Holland, of course. The machine analyzed 346 Rembrandt paintings for subject matter, lighting, color, brish technique, and so on, and created a completely new picture that looks very Rembrandty. It’s a portrait of a man (an imaginary, statistically average man) dressed and groomed in the style of the period, lit in Rembrandt’s way, executed in Rembrandt-style brush strokes (done with a 3-D printer).

Matissifier (Techxplore)

DeepArt is a much simpler affair. It takes a photo and re-renders it with the color and brush clichés of the designated artist. Fun, if you want to see how Van Gogh would have painted Marilyn Monroe or something, but not as original as a whole new pseudo-Rembrandt.

“I’m sorry, Dave, I’m afraid I can’t do that.” (Phys.org)

Everyone talks about the First Law of Robotics, but the Second Law says, “A robot must obey all orders given it by a human being, unless…” The “unless” is very important. This article discusses when a robot should disobey. Implementing proper disobedience depends on methods that sound a lot like the “consequence engine” Jim discusses elsewhere.

The Turing Tour (Phys.org)

The University of Southampton, in England, is running tests to see if people can tell driveless cars from driven ones by the way they act — an automotive version of the Turing Test.

Toy therapist (Phys.org)

There aren’t enough physical therapists to go around. These toy-sized NAO robots aren’t therapists, but they can coach and encourage you through your exercises in the therapist’s absence.