“Interpretability Will Not Reliably Find Deceptive AI” by Neel Nanda

by LessWrong (Curated & Popular)