John Schulman

American AI researcher — co-founder of OpenAI — who invented Proximal Policy Optimization (PPO) and co-created RLHF (Reinforcement Learning from Human Feedback) — the training technique behind ChatGPT — who left OpenAI in 2024 to join Anthropic.

Co-creates RLHF as the technique that made LLMs useful — Schulman's work on using human feedback to fine-tune language models (rather than just pretraining on data) produced the ChatGPT capability to follow instructions helpfully; every major useful LLM uses variants of his RLHF approach.