Buck comments on Scalable Oversight and Weak-to-Strong Generalization: Compatible approaches to the same problem

[Buck]

Jan Leike makes some very similar points in his post here.