Google DeepMind, fresh off a Nobel Prize for its AlphaFold protein-structure breakthrough, is racing to prove it can turn AI into more scientific wins while meeting commercial demands. CEO Demis Hassabis is pushing projects from genomics and materials discovery to weather and fusion, even as the rise of large language models intensifies competition and forces faster product cycles for Gemini. The company’s science-led culture and safety committees aim to blunt ethical risks, but some staff bristle at a more commercial tilt under Google. With rivals like OpenAI and Mistral launching science efforts, DeepMind’s challenge is to replicate AlphaFold’s impact in tougher domains where data are scarcer and success metrics fuzzier—without compromising on responsibility.
Related articles:
Highly accurate protein structure prediction with AlphaFold
Magnetic control of tokamak plasmas through deep reinforcement learning
Skillful nowcasting of precipitation using deep generative models of radar
Discovering faster sorting algorithms using deep reinforcement learning
What AlphaFold means for structural biologists
