Joseph Miller
Bio
Joseph Miller is an AI safety researcher based in London, UK, focused on mechanistic interpretability and AI alignment. He studied Computer Science at the University of Warwick (2017-2020) and subsequently worked as a Research Engineer at FAR.AI (Foundational AI Research), where he co-authored work on adversarial robustness and transformer circuit evaluation. As a MATS (ML Alignment & Theory Scholars) Fellow in 2024, he co-developed gradient routing, a training method that applies data-dependent gradient masks during backpropagation to isolate capabilities within specific neural network subregions, enabling interpretable representations and robust unlearning. He is also a co-author of the 2025 paper "Open Problems in Mechanistic Interpretability" and is active on LessWrong and the EA Forum. Beyond technical research, he leads PauseAI UK as its national director, advocating for a pause on advanced AI development until safety can be ensured.
Links
- Personal Website
- https://josephmiller.xyz/
- Twitter / X
- -
- LessWrong
- josephm
Grants
No grants recorded.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Mar 22, 2026, 10:29 PM UTC
- Created
- Mar 20, 2026, 3:00 AM UTC