Lei M. Zhang
banner
l32zhang.bsky.social
Lei M. Zhang
@l32zhang.bsky.social
Research Scientist @ Google DeepMind. Previously @ OpenAI. Building AGI. 🤖
Reposted by Lei M. Zhang
Reader, I do not know the demons that haunt your dreams, the frustrations and insults of life that have brought you to your knees, the sadness you battle in your heart.

But know this: there is pure goodness in this world. It exists. Sometime we see a glimpse of it.

Today, I got such a peek.
November 17, 2024 at 5:45 PM
We tried something similar in adaptive agents arxiv.org/abs/2301.07608 (multiturn adaptation + RL). This was with transformers but not LLMs. Someone should try this with LLMs!
Human-Timescale Adaptation in an Open-Ended Task Space
Foundation models have shown impressive adaptation and scalability in supervised and self-supervised learning problems, but so far these successes have not fully translated to reinforcement learning (...
arxiv.org
November 21, 2024 at 2:20 AM
Reposted by Lei M. Zhang
now, if we think of p(output | prompt, a few examples) as a predictive distribution p(y|x, D) ... it looks very much like learning to me :)

see e.g. my slide deck on drive.google.com/file/d/1B-Ka...
November 21, 2024 at 12:43 AM