Laura Weidinger: Ethical Risks, Harms, and Alignment of Large Language Models
The Gradient: Perspectives on AI
00:00
AI Ethics - You Can't Explain Deep Learning?
I think it's probably a safe assumption that these policy research teams are looking at this AI ethics type of work. And on the other side, you also mentioned part of the goal is how to mitigate the harms and that we don't have enough mechanisms for that. The paper was really driven by the technical safety team at DeepMind. So there is a team that is looking at reward, misspecification, for example, technical misalignment. We got involved from the ethics point of view as well.
Transcript
Play full episode
