Geoffrey Irving comments on How do you talk about AI safety?