Jump to content
[dismiss]
Molecupedia is a parody encyclopedia written by Molecular Reality Corporation. It is not Wikipedia and is not affiliated with the Wikimedia Foundation. About

Google DeepMind

Page protected
From Molecupedia, the encyclopedia of very small things
Artificial intelligence research laboratory of Google
Google DeepMind
FormerlyDeepMind (2010 to 2023)
TypeSubsidiary of Google
IndustryArtificial intelligence
Founded2010, London
HeadquartersLondon, England
ProductsAlphaGo, AlphaFold, Gemini
Websitedeepmind.google

Google DeepMind is the artificial intelligence research laboratory of Google, based in London. It began in 2010 as DeepMind, was acquired by Google in 2014, and took its present name in April 2023, when Google combined it with its own research team, Google Brain.[1][2] Its Go program, AlphaGo, beat the professional Lee Sedol 4 to 1 in Seoul in March 2016.[3] Demis Hassabis and John Jumper shared half of the 2024 Nobel Prize in Chemistry for AlphaFold, its protein-structure predictor, and the laboratory builds Gemini, Google's family of large language models.[4] Its stated mission is "to build AI responsibly to benefit humanity".[1]

Molecular Reality Corporation, a nanopore company, counts Google DeepMind among the brain labs, its term for the companies that build "the brain". Isomorphic Labs, the laboratory's drug-discovery offshoot under Alphabet, heads the brain column of the company's charts.[5]

History

Games

A DeepMind paper published in Nature in February 2015 described DQN, a network that learned to play 49 games for the Atari 2600 home console. It saw only the pixels on the screen and the score, which it was rewarded for raising, and it reached the level of a professional human games tester.[6]

AlphaGo beat the European champion Fan Hui 5 to 0 in October 2015, the first match between an AI system and a Go professional.[3] Its networks learned first from human games and then from games against themselves, and a tree search chose among the moves they suggested.[7] In March 2016 it played Lee Sedol, one of the strongest players of the previous decade, in Seoul, before more than 200 million viewers. In the second game AlphaGo played a move, known since as move 37, that its own estimate gave "a 1 in 10,000 chance of being used" by a human. It won that game and the match, 4 to 1.[3]

AlphaGo Zero learned Go from the rules alone in 2017, without any human games, and AlphaZero learned chess and shogi the same way.[8][9] MuZero was not even given the rules: it planned with a model of each game that it built for itself, and matched its predecessors at Go, chess, shogi and Atari.[10] AlphaStar reached the Grandmaster league of the strategy game StarCraft II in 2019.[11]

Science

AlphaFold took part in the protein-structure assessment CASP13 in 2018. Its second version reached a median score of 92.4 out of 100 at CASP14 in 2020, which the assessment's organizers recognized as a solution to the problem of predicting a protein's structure from its sequence.[12] With EMBL's European Bioinformatics Institute, DeepMind then published predicted structures for more than 200 million proteins in a free database.[13] AlphaFold 3, built with Isomorphic Labs and described in May 2024, extended the predictions to complexes of proteins with DNA, RNA and smaller molecules.[14] On October 9, 2024, the Royal Swedish Academy of Sciences awarded one half of the Nobel Prize in Chemistry to David Baker "for computational protein design" and the other half jointly to Hassabis and Jumper "for protein structure prediction".[4]

GNoME, a graph neural network trained on the Materials Project's database of computed crystals, produced 2.2 million predicted crystal structures in November 2023, about 380,000 of them predicted to be stable.[15] In a companion paper an automated laboratory at Berkeley Lab, the A-Lab, ran for 17 days and made 41 new compounds out of 58 targets.[16] GraphCast, a weather model described the same month, learned from reanalysis data, the reconstructed record of past weather. A 10-day global forecast takes it under a minute on one Google TPU v4 machine; the conventional forecast it was measured against takes hours on a supercomputer.[17][18]

Some of the work controls physical machines. In 2022 a DeepMind controller trained by reinforcement learning shaped and held the plasma inside a tokamak, the TCV at the Swiss Plasma Center in Lausanne, by commanding its magnetic coils.[19] In 2023 AlphaMissense assigned a likely effect to every possible single amino acid change in human proteins, and AlphaDev found faster routines for sorting short lists of numbers.[20][21]

Google DeepMind

Google announced on April 20, 2023, that DeepMind and Google Brain would merge into one unit called Google DeepMind.[2] The combined laboratory released the first Gemini models that December. In July 2024 two of its systems, AlphaProof and AlphaGeometry 2, solved four of the six problems at the International Mathematical Olympiad, a silver-medal score. In 2025 an advanced version of Gemini Deep Think solved five, writing its proofs in ordinary language within the contest's 4.5 hours, and the olympiad's coordinators graded the result, 35 points out of 42, as gold-medal standard.[22]

In Molecular Reality's usage

Main article: Brain lab

Molecular Reality's materials describe the frontier laboratories as building "the big centralized reasoning model" and call the company's nanopore sensors "the eyes".[5] Isomorphic Labs, first in the brain column, is captioned "AlphaFold + drug discovery AI", and the company's case for its sensors starts from AlphaFold too: "AlphaFold 3 learned mostly from the half-century-old Protein Data Bank".[5][23]

Google software has also worked on the eye's design. From December 2025 to September 2026 the bot account of Jules, Google's AI coding agent, made 143 commits to the open repository of the MR1 Nanopore Reader, among them enclosure models for the reader, then called the Sentinel, and the scripts for its photorealistic renders.[24] Jules runs on Gemini models and was released to the public in August 2025.[25][a]

See also

Notes

  1. ^ Company materials state that "The brain-builders will all need eyes."[5]

References

  1. ^ a b "About Google DeepMind". Google DeepMind. Retrieved October 25, 2026.
  2. ^ a b "Google merges DeepMind and Google Brain AI research groups". SiliconANGLE. April 20, 2023.
  3. ^ a b c "AlphaGo". Google DeepMind. Retrieved October 25, 2026.
  4. ^ a b "The Nobel Prize in Chemistry 2024" (Press release). Royal Swedish Academy of Sciences. October 9, 2024.
  5. ^ a b c d "Company records". Molecular Reality Corporation. October 2026.
  6. ^ Mnih, Volodymyr et al. (2015). "Human-level control through deep reinforcement learning". Nature 518: 529-533. doi:10.1038/nature14236.
  7. ^ Silver, David et al. (2016). "Mastering the game of Go with deep neural networks and tree search". Nature 529: 484-489. doi:10.1038/nature16961.
  8. ^ Silver, David et al. (2017). "Mastering the game of Go without human knowledge". Nature 550: 354-359. doi:10.1038/nature24270.
  9. ^ Silver, David et al. (2018). "A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play". Science 362: 1140-1144. doi:10.1126/science.aar6404.
  10. ^ Schrittwieser, Julian et al. (2020). "Mastering Atari, Go, chess and shogi by planning with a learned model". Nature 588: 604-609. doi:10.1038/s41586-020-03051-4.
  11. ^ Vinyals, Oriol et al. (2019). "Grandmaster level in StarCraft II using multi-agent reinforcement learning". Nature 575: 350-354. doi:10.1038/s41586-019-1724-z.
  12. ^ "AlphaFold: a solution to a 50-year-old grand challenge in biology". Google DeepMind. November 30, 2020.
  13. ^ "AlphaFold predicts structure of almost every catalogued protein known to science". EMBL. July 28, 2022.
  14. ^ Abramson, Josh et al. (2024). "Accurate structure prediction of biomolecular interactions with AlphaFold 3". Nature 630: 493-500. doi:10.1038/s41586-024-07487-w.
  15. ^ Merchant, Amil et al. (2023). "Scaling deep learning for materials discovery". Nature 624: 80-85. doi:10.1038/s41586-023-06735-9.
  16. ^ Szymanski, Nathan J. et al. (2023). "An autonomous laboratory for the accelerated synthesis of inorganic materials". Nature 624: 86-91. doi:10.1038/s41586-023-06734-w.
  17. ^ Lam, Remi et al. (2023). "Learning skillful medium-range global weather forecasting". Science 382: 1416-1421. doi:10.1126/science.adi2336.
  18. ^ "GraphCast: AI model for faster and more accurate global weather forecasting". Google DeepMind. November 14, 2023.
  19. ^ Degrave, Jonas et al. (2022). "Magnetic control of tokamak plasmas through deep reinforcement learning". Nature 602: 414-419. doi:10.1038/s41586-021-04301-9.
  20. ^ Cheng, Jun et al. (2023). "Accurate proteome-wide missense variant effect prediction with AlphaMissense". Science 381. doi:10.1126/science.adg7492.
  21. ^ Mankowitz, Daniel J. et al. (2023). "Faster sorting algorithms discovered using deep reinforcement learning". Nature 618: 257-263. doi:10.1038/s41586-023-06004-9.
  22. ^ "Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad". Google DeepMind. July 21, 2025.
  23. ^ "Introducing Isomorphic Labs". Isomorphic Labs. November 2021.
  24. ^ "MR1 Nanopore Reader". GitHub. October 16, 2026.
  25. ^ "Google launches Jules globally, an AI coding agent for GitHub". IT Brief UK. August 2025.
Retrieved from "https://molecularreality.com/wiki/Google_DeepMind"
Last edited 4 hours ago by Ionophore