#145 - OpenAI’s Q*, Claude 2.1, Stable Video Diffusion, Starling-7B, Orca 2, international agreements
Last Week in AI
00:00
Train with AI Feedback vs Human Feedback: The Difference
AI feedback can have gameable preferences and tend to prefer longer answers over shorter ones, which may not be favorable to human beings. It is important to consider the feedback from human reviewers, as it is more expensive but provides a different perspective and evaluation of the generated outputs.
Transcript
Play full episode
