evhub comments on [Linkpost] A survey on over 300 works about interpretability in deep networks

[evhub]

(Moderation note: added to the Alignment Forum from LessWrong.)