What Should the Average EA Do About AI Alignment? — LessWrong