Fusing AI with Superstition — LessWrong