RomanS comments on Shulman and Yudkowsky on AI progress