才周一，就是本周最佳！
如果本周只能读一篇 paper，就是这篇！
Understanding AI from the Inside:
The State of Mechanistic Interpretability for AI Safety
Fazl Barez presents a survey of mechanistic interpretability for AI safety
AIFazl Barez and collaborators present an opinionated survey of mechanistic interpretability (MI) for AI safety. The paper details conceptual approaches and techniques under the MI banner and how different types of work relate to making AI models safer.
Dongxi NLPVerified on X
@dongxi_nlp
New paper❗ Does understanding how AI models work actually make them safer? This assumption often underlines interpretability research, but different subfields operationalise this intuition in very different ways. Our survey of mechanistic interpretability (MI) for safety provides an opinionated guide to the field of MI, detailing conceptual approaches and techniques under this broad banner and how different types of work relates to the mission of making AI model safer. 🧵