Loki zen comments on Symbol/​Referent Confusions in Language Model Alignment Experiments

[Loki zen]

Good post, and additional points for not phrasing everything in programmer terms when you didn’t need to.