Rohin Shah comments on The limits of AI safety via debate