Chercheurs Clés en AI Alignment
Les chercheurs les plus influents dans le domaine de l'alignement de l'IA
Beginner
Chercheurs Clés
Eliezer Yudkowsky
Organisation: MIRI (co-fondateur)
Contributions:
- Sequences (rationalité, AI risk)
- CEV (Coherent Extrapolated Volition)
- Intelligence Explosion
- AGI Ruin: A List of Lethalities
P(doom): ~99%
Style: Conceptual, big picture, pessimiste
Ressources: https://www.lesswrong.com/users/eliezer_yudkowsky
Paul Christiano
Organisation: ARC (fondateur), ex-OpenAI
Contributions:
- Iterated Amplification
- Debate
- ELK
- RLHF (contributions)
P(doom): ~50-70%
Style: Technique, pragmatique, solutions-oriented
Ressources: https://ai-alignment.com/
Nate Soares
Organisation: MIRI (directeur exécutif)
Contributions:
- Agent foundations
- Corrigibility research
- Value learning
P(doom): ~90-95%
Ressources: https://www.lesswrong.com/users/so8res
Evan Hubinger
Organisation: Anthropic
Contributions:
- Risks from Learned Optimization (paper clé)
- Mesa-optimization, deceptive alignment (concepts)
- Sleeper Agents
Ressources: https://www.alignmentforum.org/users/evhub
Stuart Russell
Organisation: UC Berkeley
Contributions:
- Human Compatible (livre)
- Inverse Reinforcement Learning
- CIRL (Cooperative IRL)
P(doom): ~50% (publiquement)
Style: Académique, accessible