Like coding, if you get inspired by AI for a novel idea, and can reproduce the same result independently (could code the same thing by hand) or at least understand and check every single argument (self review your code, test on your machine) and get it peer reviewed (code review, but with a real human) then I don’t see why the industry accepts the latest iteration of ChatGPT being 99% written by codex, but rejects a valid math result inspired by it.