FILE – Alphabet CEO Sundar Pichai speaks about Google DeepMind at a Google I/O event in Mountain View, Calif., Wednesday, May 10, 2023. (AP Photo/Jeff Chiu, File) Google DeepMind’s vice-president of research, Pushmeet Kohli, has publicly pushed back on the idea that the company has quietly walked away from scientific research. Writing on X, Kohli said the DeepMind Science team has not pivoted at all. It has expanded its work, he said, and is still going after the hardest problems in science and biology.The clarification did not come out of nowhere. It follows a Financial Times report that said DeepMind has broken up the team behind AlphaFold, the protein-structure prediction system that won John Jumper and Demis Hassabis the 2024 Nobel Prize in Chemistry. According to the FT, most of the original AlphaFold paper authors have been moved to other projects over the past year, many of them onto work built around Gemini. Nearly a quarter of the full-time Google DeepMind authors on those papers have left the company entirely.Kohli’s post reads like a direct rebuttal to that framing. He listed what the team is still working on: understanding protein function, decoding the genome, de-novo protein and enzyme design, plus Gemini-powered agents meant to speed up the research process itself. Then came the line that carried the most weight. The culture at Google DeepMind, he said, is to empower its scientists, who have agency over which hard goals they choose to chase. That applied to the AlphaFold team too.
What the report said about DeepMind’s AlphaFold team and its research strategy shift
The FT’s account describes a lab reorganising around large language models. Researchers who once sat on a single grand challenge have been scattered across enzyme design, nuclear fusion and genomics, with others moving to Alphabet’s drug discovery arm Isomorphic Labs, which was spun out in 2021 and now works with Novartis and Eli Lilly. Kohli himself told the paper the strategy has evolved, with DeepMind now building Gemini systems that assist scientists and eventually automate parts of the scientific method.
Why John Jumper’s move to Anthropic makes the timing of the clarification awkward for Google
The detail that stings is the talent flow. Jumper and fellow AlphaFold researcher Jonas Adler were pulled onto Google’s internal “Code Strike” team earlier this year, a group set up to close the gap with Anthropic and OpenAI on AI coding. Last month Jumper said he was leaving for Anthropic. Adler and AlphaFold colleague Alexander Pritzel are going with him.That is a Nobel laureate and two core collaborators walking out of a lab that built its identity on long-horizon science. Kohli’s argument is that scientists choosing their own problems is a feature, not a failure. Critics reading the same facts see a research lab reshaped by a chatbot race it is trying to win.