Milan Rakic

Where are you from, and what is your background?

Before joining SAIL I studied for Theoretical Physics at Durham, and then went on to complete my Physics PhD in the CMTH group at Imperial, working on supersolid and pair-density wave theory, developing models for how and why they form and analysing their collective excitations. Supersolids are a fascinating state of matter which have both superfluid and crystalline properties, leading to extra speeds of sound and other weird and wacky phenomena. Since finishing my PhD I’ve made the decision to transition into AI safety research, with a focus on evaluations.

What inspired you to stay in academia?

I was inspired to join academia for the ability to advance knowledge and work on fun and interesting problems in doing so.

What do you do in your spare time?

In my spare time I enjoy playing table tennis, reading, bouldering, cooking and hanging out with my cat Bruno.

What are your research interests?

My current research is focused on statistical estimations of worst-case outputs of models. The question I am trying to answer is “can we infer any information about the magnitude of global failures given information about smaller local failures?”, and I am tackling this problem using peaks-over-threshold theory and “probably approximately correct” frameworks.