David Lorell
Bio
David Lorell is an independent AI alignment researcher who works closely with John Wentworth on the natural abstraction research agenda. He has co-authored multiple posts and a peer-reviewed paper with Wentworth, including "Natural Latents: Latent Variables Stable Across Ontologies" (arXiv:2509.03780, 2025), which develops a mathematical framework for latent variables that remain stable across different agent ontologies. His role in the collaboration involves serving as an active intellectual sounding board — asking for clarifications, requesting examples, and probing how theoretical ideas connect to broader alignment goals — a contribution John Wentworth has credited with multiplying his research productivity severalfold. Lorell is an active participant on LessWrong and the AI Alignment Forum, where he has been a member since 2022 and has contributed posts and comments on topics including natural latents, instrumental goals, coherence theorems, and corrigibility. He has also been acknowledged for discussion by EA and alignment researchers such as Joe Carlsmith. He has received general support funding for his independent alignment research work.
Links
- Personal Website
- -
- Twitter / X
- LessWrong
- david-lorell
Grants
from Survival and Flourishing Fund
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Mar 22, 2026, 3:34 PM UTC
- Created
- Mar 19, 2026, 6:22 PM UTC