Alexander Saeri comments on Towards evidence gap-maps for AI safety