Topic: lab:deep-mind-ai

← Back to home

1. Solving olympiad geometry without human demonstrations

   Paper • Jan 17, 2024 • Nature • Trieu Trinh, Yuhuai Wu, Quoc V. Le, He He, Thang Luong

   Proving mathematical theorems requires deep logical reasoning and intuition, representing a grand challenge for AI. We introduce AlphaGeometry, a neuro-symbolic system that solves complex geometry ...

2. Accurate structure prediction of biomolecular interactions with AlphaFold 3

   Paper • May 8, 2024 • Nature • Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, John Jumper

   The interactions of proteins with other molecules, such as DNA, RNA, ligands and ions, are fundamental to biological function. Here we describe AlphaFold 3, a model that can predict the joint three...

3. Training Compute-Optimal Large Language Models

   Paper • Mar 29, 2022 • arXiv • Jordan Hoffmann, Sebastian Borgeaud, Arthur Mensch, Elena Buchatskaya, Laurent Sifre

   We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly under...

4. Mastering the game of Go with deep neural networks and tree search

   Paper • Jan 27, 2016 • Nature • David Silver, Aja Huang, Chris J. Maddison, Arthur Guez, Demis Hassabis

   The game of Go has long been viewed as the most challenging of classic games for artificial intelligence owing to its enormous search space and the difficulty of evaluating board positions and move...

5. RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control

   Paper • Jul 28, 2023 • arXiv • Anthony Brohan, Noah Brown, Justice Carbajal, Yevgen Chebotar, Xi Chen, Krzysztof Choromanski, Tianli Ding, Danny Driess, Avinava Dubey, Chelsea Finn, Keerthana Gopalakrishnan, Karol Hausman, Alex Irpan, Google DeepMind

   We introduce Robotic Transformer 2 (RT-2), a novel Vision-Language-Action (VLA) model that learns from both vast web datasets and specialized robotics data. We show that high-capacity vision-langua...

6. RoboCat: A Self-Improving Foundation Agent for Robotic Manipulation

   Paper • Jun 20, 2023 • arXiv • Konstantinos Bousmalis, Giulia Vezzani, Dushyant Rao, Coline Devin, Alex X. Lee, Maria Bauza, Todor Davchev, Yuxiang Zhou, DeepMind Robotics Team

   Creating general-purpose robots requires models that can rapidly adapt to new tasks and new physical embodiments. We present RoboCat, a self-improving foundation agent for robotic manipulation. Rob...

7. Accurate variant effect prediction using AlphaMissense

   Paper • Sep 19, 2023 • Science • Jun Cheng, Guido Novati, Joshua Pan, Clare Bycroft, Akvilė Žemgulytė, Taylor Applebaum, Alexander Pritzel, L.H. Wong, Michal Zielinski, Tobias Sargeant, Richard G. Schneider, Andrew W. Senior, John Jumper, Demis Hassabis, Pushmeet Kohli, Žiga Avsec

   Missense variants in the human genome can cause devastating genetic diseases, yet the vast majority of these variants remain clinically classified as 'variants of uncertain significance' (VUS). Bui...

8. Learning skillful medium-range global weather forecasting

   Paper • Nov 14, 2023 • Science • Remi Lam, Alvaro Sanchez-Gonzalez, Matthew Willson, Peter Wirnsberger, Meire Fortunato, Alexander Alet, Suman Ravuri, Timo Ewalds, Zachary Eaton-Rosen, Weihua Hu, Alexander Merose, Stephan Hoyer, George Holland, Jacklynn Stott, Oriol Vinyals, Shakir Mohamed, Peter Battaglia

   Global medium-range weather forecasting has long been dominated by massive, compute-intensive numerical weather prediction (NWP) models governed by atmospheric physics equations. We present GraphCa...

9. Perceiver: General Perception with Iterative Attention

   Paper • Mar 4, 2021 • arXiv • Andrew Jaegle, Felix Gimeno, Andrew Brock, Oriol Vinyals, Andrew Zisserman, Joao Carreira

   Biological systems perceive the world by simultaneously processing high-dimensional inputs from modalities as diverse as vision, audition, and touch. We introduce the Perceiver, an architecture tha...

10. IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures

    Paper • Feb 9, 2018 • arXiv • Lasse Espeholt, Hubert Soyer, Remi Munos, Karen Simonyan, Volodymyr Mnih, Tom Ward, Yotam Doron, Vlad Firoiu, Tim Harley, Iain Dunning, Shane Legg, Koray Kavukcuoglu

    Scaling reinforcement learning algorithms to utilize thousands of machines efficiently is crucial for tackling complex, visually rich environments. We introduce IMPALA (Importance Weighted Actor-Le...

11. Neural Discrete Representation Learning

    Paper • Nov 2, 2017 • arXiv • Aaron van den Oord, Oriol Vinyals, Koray Kavukcuoglu

    Learning useful representations without supervision remains a key challenge in machine learning. We propose the Vector Quantised-Variational AutoEncoder (VQ-VAE), a simple yet powerful generative m...

12. Hybrid computing using a neural network with dynamic external memory

    Paper • Oct 12, 2016 • Nature • Alex Graves, Greg Wayne, Malcolm Reynolds, Tim Harley, Ivo Danihelka, Agnieszka Grabska-Barwińska, Sergio Gómez Colmenarejo, Edward Grefenstette, Tiago Ramalho, John Agapiou, Adrià Puigdomènech Badia, Karl Moritz Hermann, Yori Zwols, Georg Ostrovski, Adam Cain, Helen King, Christopher Summerfield, Phil Blunsom, Koray Kavukcuoglu, Demis Hassabis

    Artificial neural networks excel at sensory processing and pattern recognition but struggle with the systematic and reliable execution of algorithmic tasks. We introduce the Differentiable Neural C...

13. Mathematical discoveries from program search with large language models

    Paper • Dec 14, 2023 • Nature • Bernardino Romera-Paredes, Mohammadamin Barekatain, Alexander Novikov, Matej Balog, Pushmeet Kohli

    Large language models (LLMs) have demonstrated impressive capabilities in code generation, but their ability to discover novel mathematical knowledge has been limited by hallucinations and lack of ...

14. Mastering Atari, Go, chess and shogi by planning with a learned model

    Paper • Dec 23, 2020 • Nature • Julian Schrittwieser, Ioannis Antonoglou, Thomas Hubert, Karen Simonyan, Laurent Sifre, David Silver

    Constructing agents with planning capabilities has long been one of the main challenges in the pursuit of artificial intelligence. Tree-based planning methods have enjoyed huge success in challengi...

15. Flamingo: a Visual Language Model for Few-Shot Learning

    Paper • Apr 28, 2022 • arXiv • Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Karen Simonyan

    Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family ...

16. A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play

    Paper • Dec 7, 2018 • Science • David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Demis Hassabis

    The game of chess is the most widely-studied domain in the history of artificial intelligence. The strongest programs are based on a combination of sophisticated search techniques, domain-specific ...

17. Human-level control through deep reinforcement learning

    Paper • Feb 26, 2015 • Nature • Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A. Rusu, Joel Veness, Demis Hassabis

    We present the first deep learning model to successfully learn control policies directly from high-dimensional sensory input using reinforcement learning. The model is a convolutional neural networ...

18. Gemma: Open Models Based on Gemini Research and Technology

    Paper • Feb 21, 2024 • arXiv • Gemma Team, Google DeepMind

    We introduce Gemma, a family of lightweight, state-of-the-art open models built from the same research and technology used to create the Gemini models. Gemma models are offered in two sizes: a 7 bi...

19. Competition-level code generation with AlphaCode

    Paper • Dec 8, 2022 • Science • Yujia Li, David Choi, Junyoung Chung, Nate Kushman, Oriol Vinyals

    Programming represents a complex problem-solving task that requires deep logic and algorithmic reasoning. We present AlphaCode, a system that writes computer programs at a competitive level. AlphaC...

20. Improving alignment of dialogue agents via targeted human judgements

    Paper • Sep 22, 2022 • arXiv • Amelia Glaese, Nat McAleese, Maja Trebacz, John Aslanides, Vlad Firoiu, Geoffrey Irving

    We present Sparrow, an information-seeking dialogue agent trained to be more helpful, correct, and harmless compared to prompted language model baselines. We train our model using reinforcement lea...