THE VIDEO SHELF · rlhf for language models
John Schulman - Reinforcement Learning from Human Feedback: Progress and Challenges
UC Berkeley EECS colloquium by OpenAI's John Schulman on reinforcement learning from human feedback. He covers why RL helps with truthfulness, hallucination, and the limits of behavior cloning.
Why HN readers recommended it
Recommended in LLM threads as an authoritative explanation from a ChatGPT architect of why models hallucinate and why imitation is not enough.
This is a summary from the source dataset, not an endorsement by Hacker News or the video creator.
Video & recommendation details
- Creator / channel
- UC Berkeley EECS
- Topic
- rlhf for language models
- Format
- talk
- Level
- intermediate
- HN recommenders in dataset
- 2
- HN mentions in dataset
- 5
- Recommendation years in dataset
- 2023–2023
These counts and years describe the included Hacker News records; they are not YouTube views or the video’s publication date.
From the original HN conversations
Excerpts from the recorded comments. Follow each source for the full comment and surrounding discussion.
Sure, here you go: [VIDEO] Summarizing: 1. Obviously you can't completely "solve" truthfulness because people disagree on what is and is not true. But you can go a long way. 2. The models do know what they don't know. Their level of uncertainty is not …
mike_hearn on Hacker News · 2023-04
In It's not only you, there's an explosion of ChatGPT on HNThere is this idea that the goal of RLHF is to make ChatGPT woke or as you put it to lobotomize it. I suspect that this is a conspiracy theory. There's a very good talk by John Schulman, chief architect of ChatGPT [0], where he …
KKKKkkkk1 on Hacker News · 2023-05
In Don Knuth Plays with ChatGPT
Summaries, classifications and recommendation records are from the HackerTube source dataset. Read about this library and its limitations.