David Silver

Overview

David Silver is a British computer scientist born in 1976 who specialises in reinforcement learning.^[1] Reinforcement learning is the branch of machine learning in which a program is not handed the correct answer for every example; it acts, receives a reward signal, and changes its behaviour so as to collect more reward over time. Silver is best known as the leader of the AlphaGo project at DeepMind, the Go-playing system that made learning-based game playing visible to a wide public.^[1]

Education and early career

Silver studied at Christ's College, Cambridge, and graduated in 1997.^[1] At Cambridge he became friends with Demis Hassabis.^[1] After graduating, Silver co-founded the video-game company Elixir Studios.^[1]

He went on to the University of Alberta, where he earned a PhD in 2009 with a thesis on reinforcement learning and simulation-based search in computer Go.^[1] Simulation-based search means the program plays a position forward many times inside the computer and uses the results of those trial games to judge which move to make, instead of relying only on a hand-written rule for how good a position looks. Go, with its enormous number of possible positions, was a natural place to try such a method.

Deep reinforcement learning and Atari

In 2013 Silver co-authored "Playing Atari with Deep Reinforcement Learning", the deep Q-network paper.^[2] The title states the setting: Atari games, deep reinforcement learning. A deep Q-network puts a deep neural network inside the reinforcement-learning loop, so that the network is trained by the signals that come out of the game and its output is then used to choose the next action.

AlphaGo, AlphaGo Zero and AlphaZero

Silver led the AlphaGo project and the development of AlphaZero, and he is listed first among the authors of the 2016 Nature paper on AlphaGo.^[1] The system played the Go player Lee Sedol, and the version that defeated him became the reference point for what came after.^[4]

AlphaGo Zero, as described on DeepMind's blog, took another route: it learned by playing against itself starting from random play, and it beat the AlphaGo version that had defeated Lee Sedol by 100 games to 0.^[4] Silver also led the development of AlphaZero.^[1]

DeepMind's published work continued in this direction with MuZero, which according to the company's blog learns a model of a game's dynamics and uses that model to plan, mastering chess, Go, shogi and Atari without being told the rules.^[5]

Recognition and teaching

Silver received the 2019 ACM Prize in Computing.^[1] The ACM's press release on the award describes it as going to the developer of AlphaGo, for breakthrough advances in computer game-playing.^[3] He has also been a professor at University College London.^[1]

Ineffable Intelligence

Silver worked full-time at Google DeepMind from 2013.^[1] He founded a company, Ineffable Intelligence, in November 2025, and he left Google DeepMind in January 2026.^[1] In April 2026 the company reportedly raised $1.1 billion at a valuation of $5.1 billion.^[1]

Why it matters

Read end to end, Silver's public career runs from a doctoral thesis on search and learning in computer Go to leading the development of systems that played a top human Go player in public.^[1][4] For a general audience the AlphaGo games against Lee Sedol were the most visible appearance of this line of work, and the ACM award recognised the game-playing systems specifically.^[3] AlphaGo Zero added a further point: a system that starts from random play could still beat the version that had defeated a top human player, 100 games to 0.^[4] In DeepMind's account, MuZero carries the same ingredients of learning and planning into games whose rules are not supplied.^[5] Silver's move from Google DeepMind to a company of his own changes where that research is carried out.^[1]