
4 - Risks from Learned Optimization with Evan Hubinger
AXRP - the AI X-risk Research Podcast
00:00
Do You Get Deceptive Alignment?
Deceptive alignment is the thing where your miss oftemeiser has a different goal than a base objective. So therefore, in situations where the messa objective and the base objective might come apart, if your ively lined, your mess optimiser picks the thing that the base objective wants. And so at what point do you get an objective that cares about what happens like ofter episodes? How does that end up getting selected for? Yes, so this is a really good question, and it's a complicated one. I'll try to sort o talk, talk through some of the ways in which i vow this.
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.