Rohin Shah comments on How do you talk about AI safety?