BERI-SRL Collaboration

The BERI-SRL Collaboration is a partnership between the Berkeley Existential Risk Initiative (BERI) and the Safe Robotics Laboratory (SRL) at Princeton University. SRL is directed by Jaime Fernandez Fisac, an Assistant Professor of Electrical and Computer Engineering at Princeton, whose research integrates control systems, game theory, and artificial intelligence to develop rigorous safety frameworks with clear-cut guarantees for autonomous systems. SRL's research spans several areas relevant to AI safety: safe robot learning (enabling robots to acquire new skills without risking accidents), safe human-robot interaction through game-theoretic approaches, reinforcement learning for safety in uncertain or hostile environments, and human-AI safety more broadly. Fisac has argued that AI systems need not achieve superintelligent capability to pose catastrophic risk to society, pointing to content recommendation algorithms as evidence of the destabilizing power of large-scale socio-technical feedback loops. His group has also researched AI persuasion risks, creating the Convince Me benchmark for measuring language models' ability to persuade. BERI launched a trial collaboration with SRL in 2021 as part of its model of supporting university research groups working on existential risk reduction. BERI provides operational support that is difficult or slow to obtain through university administration, including hiring research staff, purchasing equipment and resources, and handling administrative tasks. In December 2023, BERI announced that the trial collaboration had been converted into a main collaboration, supported by a $52,000 grant from Jaan Tallinn through the Survival and Flourishing Fund's 2023 H2 grant round. The SRL team at Princeton includes approximately 14 members: the PI, eight PhD students, one visiting student, and four undergraduate researchers. The lab's work has been recognized through publications at top venues including NeurIPS and RSS, and Fisac has received a Sony Focused Research Award, an NSF CAREER Award, and Google faculty research awards.

Endorsements support Berkeley Existential Risk Initiative.

No endorsements yet.

Discussion

Start Date

-

End Date

-

Expected Duration

-

Funding Raised to Date

-