DeepMind disbands its Nobel Prize-winning AlphaFold team.

      Less than a year ago, AlphaFold earned Google DeepMind a Nobel Prize. Now, DeepMind has disbanded the team that developed it. According to the Financial Times, the lab has reassigned most of the authors of the original AlphaFold paper over the past year, with nearly a quarter leaving the company entirely. DeepMind has confirmed these changes and is integrating the work into a broader initiative surrounding Gemini.

      From tackling a grand challenge to becoming the ‘AI scientist’

      AlphaFold represented the clearest manifestation of DeepMind’s previous approach: dedicate a team to a significant challenge and resolve it. As noted by Engadget, it solved the 50-year protein-folding dilemma, accurately predicting 3D protein structures in minutes. Its technology now supports drug discovery and disease research globally.

      That approach is now being phased out. DeepMind's research VP, Pushmeet Kohli, told the FT, “Our strategy over the last nine years has been to focus on grand challenges. The strategy has evolved.” Instead of having one team for each problem, the lab is creating Gemini-powered systems designed to aid scientists and automate aspects of research itself.

      Former AlphaFold team members have dispersed. Some have transitioned to Gemini projects, while others have moved into areas such as enzyme design, nuclear fusion, and genomics, as reported by The Decoder. A few have joined Isomorphic Labs, the Alphabet subsidiary in drug discovery inspired by AlphaFold.

      Major departures to Anthropic

      The most significant losses have departed entirely. John Jumper, who shared the 2024 Nobel Prize with DeepMind chief Demis Hassabis, transitioned to Anthropic in June. Two principal AlphaFold researchers, Jonas Adler and Alexander Pritzel, followed him.

      A DeepMind employee mentioned to the FT that these three were “instrumental, important, core members” whose departures “caught the team by surprise.” They are now at a competitor that recently launched Claude Science, a platform tailored for the very biology and drug-discovery tasks pioneered by AlphaFold.

      The significance of this shift

      One optimistic interpretation is that AlphaFold simply fulfilled its purpose. It solved its problem, led to the creation of a startup, and allowed its team to pursue new endeavors. DeepMind expresses continued pride in the work.

      However, a more concerning perspective is what this transition indicates. The lab, once characterized by profound scientific exploration, is reorganizing to compete in the same language model race as its peers. Moreover, it is losing the talent that established its reputation to rivals advancing that competition. Betting on an “AI scientist” is a broader, less defined goal compared to solving a protein-folding issue. Whether this will yield results similar to those of AlphaFold remains uncertain.