AXRP - the AI X-risk Research Podcast cover image

4 - Risks from Learned Optimization with Evan Hubinger

AXRP - the AI X-risk Research Podcast

00:00

Is There a Behaviour Objective?

A thermistat is an example of a system that has a behavior objective, but not necessarily a maso objective or an internally represented objective. So i want to ask another question about this e, this sort of interplay between this idea of like, am planning and internally represented objectives. If you're trying to look at a trained model and understand what that model tive is, you you have to like, have transparency tools, or some means of being able to look inside the model that lie gives you a real understanding of what is doing. You can imagine tha situation. I think clearly, the answer is, well, the mase objective should be the, you know, the negative

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app