zac_kenton comments on On scalable oversight with weak LLMs judging strong LLMs

[zac_kenton]

The post has now been edited with the updated plots for open consultancy/​debate.