4 - Risks from Learned Optimization with Evan Hubinger
AXRP - the AI X-risk Research Podcast
00:00
Is There a Way to Get a Masoprex?
E shek: What types of learning tasks do you think are more or less prone to producing demisers? Ye, that's a great question. So we do talk about this litl bit in the paper. E shek: The r l qrestiunis is a little bit tricky,. so we e definitly mostly focus on r l. I do think tyou can get mas oponization, answe t more supervised learning settings. But see Ma, you'll serve with the other question though, just just around the environment. You know, there's a natural way in which sort of opprnization can be nice in that situation.
Transcript
Play full episode
