Don't Dismiss Simple Alignment Approaches — LessWrong