Demis Hassabis

In a December 5, 2019 retrospective, Hassabis describes starting DeepMind nearly ten years earlier with a mission to understand and recreate intelligence. His research overview discusses protein prediction, cooperating game agents and StarCraft II.^[1]

Read this as a participant’s account of institution building and research ambitions. It does not independently certify the impact of those programs or establish safe general intelligence. The origins guide distinguishes that capability agenda from communicating purposes and retaining human control.

For a supervision method from that research ecosystem, read Deep reinforcement learning from human preferences and inspect its own author list and experiments. Conceptual proximity does not establish Hassabis’s authorship of that paper.

Review scope: selected opening and research overview in the dated original retrospective; linked experiments and subsequent organizational roles not audited.