AlphaGo is showing that it can come up with moves that no human ever thought they could work; even the previous versions that used human matches for learning was able to do that, and now AlphaGo Zero is doing it from scratch. The Go world is very excited and Pro players are being inspired by these new moves. The scale is very different to the examples you mentioned, though. AlphaGo has played millions of matches to be able to achieve brilliance; The Next Rembrandt is just one painting and it would hardly be consider creative. Maybe if it did hundreds of thousands it could start producing original work, as a Rembrandt disciple if you will.
That said, I think your concern is spot on, particularly in marketing and advertising. It’s one of the reasons why we’re building Seenapse, which goes in the opposite direction, using mental associations as organizing principle. It doesn’t come up with ideas, its users do.