DeepMind has produced some of the most celebrated achievements in AI research: a program that mastered Go, a system that predicted the structures of almost all known proteins, and a Nobel Prize. Today, as Google DeepMind, it is also the engine behind Google’s Gemini models and AI products.

Founding (2010)

DeepMind was founded in London in 2010 by Demis Hassabis, a former chess prodigy, video game designer and neuroscientist; Shane Legg, a machine learning researcher focused on artificial general intelligence; and Mustafa Suleyman, an entrepreneur with a background in social policy. Its ambition was to “solve intelligence” and use it to solve other problems.

The Atari breakthrough

In 2013, DeepMind showed that a single deep reinforcement learning system, the Deep Q-Network (DQN), could learn to play many Atari games from raw pixels and scores, published in Nature in 2015. It was an early demonstration of general learning across tasks.

Acquisition by Google (2014)

Google acquired DeepMind in 2014, reportedly for around £400 million or more. DeepMind remained a distinct unit based in London, with an ethics board as part of the agreement.

AlphaGo (2016)

AlphaGo defeated Lee Sedol, one of the world’s best Go players, 4–1 in Seoul in March 2016 — a milestone many experts had expected to be at least a decade away. Its “move 37” in the second game became famous as a creative move few humans would have played. Successors AlphaGo Zero and AlphaZero learned purely from self-play, mastering Go, chess and shogi.

Science: AlphaFold and beyond

In 2020, AlphaFold 2 effectively solved the protein structure prediction problem in the CASP14 assessment. DeepMind and EMBL-EBI released predicted structures for over 200 million proteins. In 2024, Hassabis and John Jumper shared the Nobel Prize in Chemistry (with David Baker) for protein structure prediction.

Other scientific projects include:

  • AlphaStar, which reached Grandmaster level in StarCraft II;
  • GraphCast and later weather models outperforming traditional forecasts on many measures;
  • GNoME, which predicted hundreds of thousands of new stable crystal materials;
  • AlphaGeometry and AlphaProof, which achieved results at International Mathematical Olympiad level;
  • fusion plasma control with the Swiss Plasma Center.

Health and controversy

DeepMind’s early health work included a collaboration with the Royal Free London NHS Foundation Trust. In 2017, the UK Information Commissioner’s Office ruled that the trust had failed to comply with data protection law when sharing patient data with DeepMind for an app to detect acute kidney injury. The case became an important lesson in health data governance.

Merger with Google Brain (2023)

In April 2023, amid competition from ChatGPT, Google merged DeepMind with its Google Brain team to form Google DeepMind, led by Hassabis. The combined unit develops the Gemini family of multimodal models, first launched in December 2023, which power Google’s assistant, search features, Workspace and developer products. It also develops open-weight Gemma models, image and video generators, and robotics models.

People

Mustafa Suleyman left DeepMind in 2019, co-founded Inflection AI, and later became head of Microsoft AI. Shane Legg remains chief AGI scientist. Many DeepMind alumni have founded startups or joined other labs.

Culture and approach

DeepMind is known for a research-driven culture, long-term projects, an emphasis on science applications, and an explicit focus on AGI and its safety. Integration into Google’s product machine has increased commercial pressure and the pace of releases.

Isomorphic Labs

In 2021, Alphabet launched Isomorphic Labs, led by Demis Hassabis, to apply DeepMind’s AI methods to drug discovery. It has signed partnerships with major pharmaceutical companies and co-developed AlphaFold 3 with Google DeepMind.

What to watch next

  • New Gemini model generations and agent capabilities.
  • Scientific breakthroughs from AlphaFold successors and materials research.
  • Isomorphic Labs’ drug discovery progress.
  • DeepMind’s AGI safety research and frameworks.

Key takeaways

  • DeepMind was founded in London in 2010 by Hassabis, Legg and Suleyman.
  • DQN, AlphaGo and AlphaZero were landmark reinforcement learning achievements.
  • AlphaFold transformed biology and earned a 2024 Nobel Prize in Chemistry.
  • In 2023, DeepMind merged with Google Brain to form Google DeepMind.
  • Google DeepMind develops the Gemini models behind Google’s AI products.

Frequently asked questions

Is DeepMind part of Google?

Yes. DeepMind was acquired by Google in 2014 and is now Google DeepMind within Alphabet.

What is DeepMind best known for?

AlphaGo’s victory over Lee Sedol, AlphaFold’s protein structure predictions and, more recently, the Gemini models.

Who leads Google DeepMind?

Demis Hassabis, a co-founder, is CEO of Google DeepMind.