MS&E faculty among AI Alignment Project awardees

The AI Security Institute chose research projects involving MS&E faculty to be among the first projects awarded funding through its AI Alignment Project.

The Alignment Project supports research that helps ensure that AI systems do what humans intend, even when complexity and the stakes are high. Out of more than 800 submitted proposals, 60 were selected to receive the first round of grants.

Professor Ben Van Roy's project, A Mathematical Model of Misalignment, will formulate a general mathematical model of how a training process can give rise to a misaligned AI system that poses catastrophic risk.

Another project involves Professor Van Roy and Professor Ramesh Johari, and is led by John Duchi, Associate Professor of Statistics and Electrical Engineering. Their project, (Mis)aligned Preferences, will study what data is needed for agents to reflect honest human preferences.

Learn more about the Alignment Project, and see the full list of awardees.