
7 - Side Effects with Victoria Krakovna
AXRP - the AI X-risk Research Podcast
How to Deploy a Reference Agent on a Future Task
So what we do here is we introduce this comparison to the the base line policy, basically byby comparing to a reference agentn on this future task. And so like, for example, in the sushi case, the the reference agent would have just beter allowed the soushi to get eaten. So that, wo kno, it doesn't actually get reward for the suci related task, for example. This is sort of trying to capture this idea that you can't do better than the base line, and cherms off getting this future task reward.
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.