
23 - Mechanistic Anomaly Detection with Mark Xu
AXRP - the AI X-risk Research Podcast
The Limits of Mechanistic Anomaly Detection
So I guess we've talked about mechanistic anomaly detection a while. Yeah, maybe more broadly, like how are you thinking about on the two scales of like, it doesn't really solve this to it totally solves this. And then there's all sorts of empirical middle grounds. Or there I think worlds where we don't like nail the problem. But we like can do some forms of ad hoc mechanistic anomaly detections using normal human interpretability and being like, huh, that used looked weird. The neurons are distributed weirdly. Now, perhaps that's not good.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.