We believe that the design and deployment of safe artificial intelligence is among the most important problems of our time. We also think it is an interesting problem for scientists and policymakers alike.

AI has advanced rapidly in the last decade, exhibiting unprecedented capabilities and performance [1].

As AI systems continue to develop, they will serve a growing social, scientific, and economic function. We expect current practices to lead to highly advanced AI systems that provide enormous societal benefits. However, we also expect current practices to produce unsafe AI systems and possibly pose serious harms [2].

Fortunately, there are avenues of research and policy that can help improve the safety of AI systems, making them more robust and more rigorously considerate of the persons affected by them [3].

Our goal is to mitigate risks from advanced AI by conducting AI safety research and equipping students to pursue careers in AI safety.