> A lot of people in this thread have an extremely facile understanding of art,
I don't think anyone would deny Lynch had his own, distinct Lynch-ian stye. A lot of people probably can't describe it clearly in words but they'll recognize when they see it. So I think it's plausible LLMs "getting the gist" of the style and asking them to generate something "Lynch-ian" might just work.
I propose a Turing test for it. Show a bunch of people unfamiliar with most of his work some of his works mixed in with some generated, and see if they could tell which is which. It will fail today, but I think in 10 years we'll get there, even if not sooner.
I don't think anyone would deny Lynch had his own, distinct Lynch-ian stye. A lot of people probably can't describe it clearly in words but they'll recognize when they see it. So I think it's plausible LLMs "getting the gist" of the style and asking them to generate something "Lynch-ian" might just work.
I propose a Turing test for it. Show a bunch of people unfamiliar with most of his work some of his works mixed in with some generated, and see if they could tell which is which. It will fail today, but I think in 10 years we'll get there, even if not sooner.