Thane Ruthenis comments on Why Not Just Train For Interpretability?