Jan Leike

Machine learning & alignment researcher

Optimizing for a post-AGI future where humanity flourishes

About me

Until recently, I co-led the Superalignment Team at OpenAI, where I’ve been involved in the development of InstructGPT, ChatGPT, and the alignment of GPT-4. I developed OpenAI’s approach to alignment research and co-authored the Superalignment Team’s research roadmap. Prior to OpenAI, I was an alignment researcher at DeepMind where I prototyped reinforcement learning from human feedback. I hold a PhD in Reinforcement Learning theory from the Australian National University. In 2023 TIME magazine listed me as one of the 100 most influential people in AI.

How can we train AI systems to follow human intent on tasks that are difficult for humans to evaluate directly?

My former team at OpenAI is researching how to align an automated alignment researcher, working on scalable oversight, weak-to-strong generalization, automated interpretability, and model organisms, among other projects.

The announcement of our Superalignment Team, aiming to solve this problem by 2027, backed by large amounts of OpenAI’s compute
The podcast interview with 80,000 hours (video) is currently the best introduction into my thinking in podcast form, especially if you’re coming from machine learning.
The podcast interview with Daniel Filan goes into more technical questions, and should be interesting for those already somewhat familiar with alignment research.
My blog about alignment research.

Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision
Collin Burns, Pavel Izmailov, Jan Hendrik Kirchner, Bowen Baker, Leo Gao, Leopold Aschenbrenner, Yining Chen, Adrien Ecoffet, Manas Joglekar, Jan Leike, Ilya Sutskever, Jeff Wu. 2023.
Language models can explain neurons in language models
Steven Bills, Nick Cammarata, Dan Mossing, Henk Tillman, Leo Gao, Gabriel Goh, Ilya Sutskever, Jan Leike, Jeff Wu, William Saunders. 2023.
Self-critiquing models for assisting human evaluators
William Saunders, Catherine Yeh, Jeff Wu, Steven Bills, Long Ouyang, Jonathan Ward, Jan Leike. 2022.
Training language models to follow instructions with human feedback
Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, Ryan Lowe. Neural Information Processing Systems, 2022.
Scalable agent alignment via reward modeling: a research direction
Jan Leike, David Krueger, Tom Everitt, Miljan Martic, Vishal Maini, and Shane Legg. 2018.
Deep Reinforcement Learning from Human Preferences
Paul F Christiano, Jan Leike, Tom B Brown, Miljan Martic, Shane Legg, and Dario Amodei. Neural Information Processing Systems, 2017.
Nonparametric General Reinforcement Learning.
Jan Leike. PhD Thesis, supervised by Marcus Hutter, 2016.