Owen Cotton-Barratt comments on On Deference and Yudkowsky’s AI Risk Estimates