AXRP - the AI X-risk Research Podcast cover image

23 - Mechanistic Anomaly Detection with Mark Xu

AXRP - the AI X-risk Research Podcast

CHAPTER

The Problem With Mechanistic Anomalies in AI

I don't know if I think like concretely about that, right? Suppose that like I've got this distribution of ways the world might be and that distribution isn't changing. Like when you're drawing that sample something's going to happen that didn't happen in the first million samples it seems to be like either you flag that as an anomaly or you tell it not to do anything different which is bad. So one explanation for why your AI doesn't freak out when it's encountered something novel is that  it's still doing the same kinds of reasoning that it learned to do during training. The things that we do want to flag as anomalies is like suppose that there's like a coup

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.
App store bannerPlay store banner