Qiaochu_Yuan comments on Self-regulation of safety in AI research