← Les avertissements
Risque d’extinctionLaboratoire de pointeLead, Alignment Science team, Anthropic2 entrées au registre
Suivre
- X@janleike
- Substackaligned.substack.com
- Site webjan.leike.name
- Wikipédiaen.wikipedia.org
Jan Leike
Chercheur en alignement qui a codirigé l’équipe de superalignement d’OpenAI, démissionnaire en mai 2024 en disant que la sécurité était passée après les produits, il travaille aujourd’hui chez Anthropic, où il maintient que la supervision de modèles surhumains reste un problème non résolu.
Fonctions
- 2021 – mai 2024Head of Alignment, OpenAICo-led the Superalignment project with Ilya Sutskever from June 2023 until his resignation.Source →
- mai 2024 – aujourd’huiLead, Alignment Science team, AnthropicTeam focused on scalable oversight, weak-to-strong generalization and automated alignment research.Source →
Au registre
22 janv. 2026 · Publication · Musings on the Alignment Problem (Substack)
Once our models are so capable that on many tasks we don't understand their actions anymore, a lot of current approaches won't work the same way.
Lead, Alignment Science team, Anthropic à l’époque
Alignment is not solved but it increasingly looks solvable →
17 mai 2024 · Publication · X
safety culture and processes have taken a backseat to shiny products
Lead, Alignment Science team, Anthropic à l’époque
From the X thread announcing his resignation from OpenAI, as quoted by MIT Technology Review.