Heather A. Cam comments on AI alignment researchers don’t (seem to) stack

[Heather A. Cam]

Yeah, basically explore-then-exploit. (I do worry that the toy model is truer IRL though...)