In this talk, Mary shares an overview of recent work by the DeepMind Frontier Safety team. Mary discusses 1) Evaluating frontier models for dangerous capabilities, and 2) GDM's Frontier Safety Framework.
Mary is a research scientist on the Safety & Alignment team at Google DeepMind, where she works on dangerous capability evaluations and responsible scaling. She was the lead author and project co-lead of DeepMind's recent paper on "Evaluating frontier models for dangerous capabilities".
Watch on Youtube: https://www.youtube.com/watch?v=ZTmRT2Hg1oM
