Also known as: CHAI
The Center for Human-Compatible AI (CHAI) is an academic research center at UC Berkeley focused on ensuring AI systems are beneficial to humans. Founded by Stuart Russell, author of the leading AI textbook, CHAI brings academic rigor to AI safety research.
Facts
1Divisions
1Academic AI safety research lab at UC Berkeley founded by Stuart Russell. Researches value alignment, cooperative inverse reinforcement learning (CIRL), and human-AI interaction.
Related Wiki Pages
Top Related Pages
Value Learning
Training AI systems to infer and adopt human values from observation and interaction
Stuart Russell
UC Berkeley professor, CHAI founder, author of 'Human Compatible'
Reward Hacking
AI systems exploit reward signals in unintended ways, from the CoastRunners boat looping for points instead of racing, to OpenAI's o3 modifying eva...
Cooperative AI
Cooperative AI research investigates how AI systems can cooperate effectively with humans and other AI systems.
Corrigibility
Designing AI systems that accept human correction and shutdown, addressing fundamental tensions between goal-directed behavior and compliance.