Wonder and The Golden AI Rule — LessWrong