Larks comments on What predictions from theoretical AI Safety research have been confirmed by empirical work?

[Larks]

Conversations people have with un-RLHF’d models.