David Silver

Silver earned bachelor’s and master’s degrees at Cambridge in 1997 and 2000, then a PhD in computer science at the University of Alberta in 2009. The ACM’s biography also records that he co-founded Elixir Studios in 1998 and served as its chief technology officer and lead programmer. His later research brought games into the study of learning algorithms, where a system can repeatedly act, observe outcomes and improve its choices. His UCL teaching makes the underlying concepts available beyond the teams building those systems.[2][5][4]

Silver’s current biography identifies him as CEO of Ineffable Intelligence, a professor at UCL and the former lead of DeepMind’s reinforcement-learning team. In a note dated January 15, 2026, he explains his decision to set up a company focused on discovering knowledge through experience in an environment. He argues for a research setting with room for ambitious work without immediate pressure for incremental products. The company’s goal of creating a superlearner is an ambition; the note explains the research direction rather than reporting that the goal has been achieved.[1][7][3]

AlphaGo combined learned judgment with search

Silver is a co-author of the 2016 AlphaGo paper. The system uses one neural network to evaluate board positions and another to select promising moves. Training combines examples from human expert games with games of self-play, then a search procedure brings those networks together. The published evaluation includes a five-to-zero victory over the European Go champion. The work is a useful example of learned judgment supporting explicit search: the networks help decide which possible futures are worth exploring and how favorable a position is.[4]

Learning the foundations

Silver’s 2015 reinforcement-learning course begins with the problem of an agent choosing actions over time and proceeds through Markov decision processes, planning, model-free learning and function approximation. Later lectures cover policy gradients, exploration and the integration of learning with planning. The accompanying Easy21 assignment gives learners a small environment in which to implement the ideas. This teaching sequence is a practical route into the mathematics behind his research, rather than a tour of model announcements.[5]

In Welcome to the Era of Experience, Silver and Richard Sutton argue for agents that learn continually from interaction, with observations, actions and reward signals grounded in an environment. They distinguish this proposed direction from relying primarily on fixed collections of human examples. The essay considers longer-lived agents and feedback about the consequences of their actions, while also acknowledging oversight and interpretability risks. It is a statement of research priorities and expectations, not proof that experience alone has produced a generally superhuman system.[6]