Gunnar_Zarncke comments on AI #33: Cool New Interpretability Paper