This should probably link to this instead: https://www.astralcodexten.com/p/ai-art-turing-test

As the twitter link just links there.

But I still don't think this is quite answering the right question. Yes, it's amazing, and by many definitions of "creative" it may not matter. If you want to have a house filled with custom artwork in any style (maybe it has to be an existing style) then it may be a distinction without a difference.

But his test doesn't seem to me to actually answer the right question. It tells you that AI can create something indistinguishable from Monet or Basquiat, but can it convince people to care about new art?

The thing that really makes me curious about this is not that I necessarily think it can't but that there's a middle ground here. You have to create something novel that people are attracted to, and be able to market it. (I think AI can do this) but the space can't be so saturated nothing rises to the top.