I feel like this is just "we lied to people and we are surprised they trusted us" which just emphasize how hard we should work to enforce generated content as such (which I don't even know if it is possible in a reliable way)

Anyway, the good thing with human mind is it can explore, where a LLM is statistically producing words. It will certainly produce average stories, which if they are short will be better than most human writing. But it won't explore new ways of writing, which are the ones which can become a masterpiece IMHO

I would be interested in long stories experiments though, is the context wide enough to keep consistency and mid quality or are they becoming boring and un-readable like claude can be when trying to explain a concept ?

Then starts the philosophical questions : If a human need to write world rules, names, locations, characters, psychology, how each character perceive the world despite its rules, define relationship between characters and then delegate the story writing to a LLM, how much of the story belong to the human and what's the difference between doing this and hiring a ghostwriter