Training impossibilities aside, how would you even make an optimization loop for "fun"?
replyBrain-computer interface, maybe. Hook a million play-testers up to the output and have it iterate.
reply..feels like there was a black mirror episode about that though