Slightly off topic but I found contemplating this possibility amusing: Would it be possible that once we manage to eliminate obvious logical and contextual mis

Slightly off topic but I found contemplating this possibility amusing: Would it be possible that once we manage to eliminate obvious logical and contextual mistakes in the generated texts, they could be used to discover alternative (and consistent) views of the world (e.g. about art, philosophy...)?. The AI would be able to create a huge number of theories and it's possible that some of them would be both interesting and original. It would be a kind of restrained (restrained because we would prune mistakes we do not want the AI to make and wouldn't just be a random typing) infinite monkeys way of exploring theories about the world. It would be even funnier if we could filter the subset of generated tests that is testable :)