A semester-long introductory technical reading group on AI safety research, covering neural network interpretability, learning from human feedback, goal misgeneralization in reinforcement learning agents, eliciting latent knowledge, and evaluating dangerous capabilities in models.

View

Free

Event

Introductory AI Policy Reading Group

Fri, 21 Aug

An introductory AI policy reading group discussing core strategic issues posed by the development of transformative AI systems.

View

Services

2 services · 1 category

AI Policy Reading Group

Technical Reading Group

About

Hours and details

Harvard AI Safety Team researches and evaluates the capabilities and risks of language-model agents in realistic, open-ended environments. Its work focuses on autonomous replication and adaptation, helping inform AI security, monitoring, and alignment.

Education
Community and government