6 - Debate and Imitative Generalization with Beth Barnes
AXRP - the AI X-risk Research Podcast
00:00
Machine Learning
Machine learning generalization is kind of scary. We have no particular reason to believe that the that sort of implicit in m l architectures is a good prior. The worry is something like, something that performs really well on your training data is an agent that reasons about the process it's embedded in and then, like, who knows what that does later on. So we're always going to have to do this generalization between questions that we know the answers to. Or like questions that we can't supervise at all.
Transcript
Play full episode
