Google DeepMind may be disbanding its AlphaFold protein structure prediction AI development team to focus on Gemini.

The Financial Times has reported that Google DeepMind has effectively disbanded the development team for AlphaFold, an AI that predicts the three-dimensional structure of proteins. Many of the researchers have moved on to research utilizing large-scale language models such as Gemini, and some have left Google, but AlphaFold itself and related services will not be discontinued.

Google DeepMind dismantles Nobel-winning AlphaFold team in strategic shift
https://www.ft.com/content/61b2953d-ee0d-45de-af6e-a9c1cf524b33?syn-25a6b1a6=1

Google Shuts Down Its Nobel-Prize Winning AlphaFold Project As It Focuses On Gemini
https://www.engadget.com/2225849/google-shuts-down-alphafold/

AlphaFold is an AI that predicts the complex three-dimensional structure of proteins from their amino acid sequences. It has enabled highly accurate prediction of protein structures, which previously required lengthy experimental analysis, and has had a significant impact on biology and drug discovery research.

DeepMind, an artificial intelligence company, has developed 'AlphaFold,' a tool that predicts the three-dimensional structure of proteins - GIGAZINE

Google DeepMind began developing AlphaFold around 2018, and with the release of AlphaFold 2 in 2020, they dramatically improved the accuracy of protein structure prediction. For this achievement, Google DeepMind CEO Demis Hassabis and John Jumper, who led the development of AlphaFold, were awarded the Nobel Prize in Chemistry in 2024.

Google DeepMind co-founder Demis Hassabis and others receive Nobel Prize in Chemistry; their research on protein structure prediction AI, including 'AlphaFold,' is highly praised - GIGAZINE

A Financial Times analysis of personnel changes revealed that the majority of researchers who authored the main papers on AlphaFold had moved to other projects in the past year. Google DeepMind has confirmed that researchers have moved to Gemini-based projects, as well as to fields such as enzyme design, genome research, and nuclear fusion.

Some researchers have moved to Isomorphic Labs, an Alphabet subsidiary that is advancing AI-powered drug discovery. The company was established in 2021, separating from Google DeepMind, to apply AlphaFold's technology to drug development, and has partnered with major pharmaceutical companies Novartis and Eli Lilly.

Meanwhile, it has been reported that about a quarter of the full-time Google DeepMind researchers who contributed to the main AlphaFold paper have already left Google. Jumper and AlphaFold researcher Jonas Adler have moved to Google's internal AI coding team, Code Strike, and have since announced their move to competitor Anthropic. Alexander Pritzel, who was involved in the development of AlphaFold, is also expected to join Anthropic.

John Jumper, who won the Nobel Prize for 'protein structure prediction,' has announced he will be leaving Google DeepMind, where he worked for about nine years, to join Anthropic - GIGAZINE

This restructuring stems from a shift in Google DeepMind's strategy for scientific research. Previously, the company formed specialized teams for each individual challenge, such as protein folding problems. However, going forward, it plans to use Gemini to support researchers and, in the future, focus on developing a general-purpose system that automates parts of scientific research.

Pushmith Kohli, who leads the AI for Science division at Google DeepMind, explains that while they have set specific 'big challenges' and driven various projects over the past nine years, their strategy has evolved. The movement to utilize large-scale language models not merely as conversational AI, but as a foundation for accelerating scientific discovery, is spreading to competitors such as OpenAI and Anthropic.

However, the dismantling of this announcement concerns the dedicated team that developed AlphaFold, and does not mean that the AlphaFold system or database itself will be shut down. Google DeepMind still presents AlphaFold as an important achievement in scientific research, and states that it is used by more than 3 million researchers worldwide. Technologies derived from AlphaFold have also been carried over to drug discovery research at Isomorphic Labs and other projects at Google DeepMind.

Google DeepMind has stated that it is proud of the global impact and scientific achievements that AlphaFold has had. This restructuring is seen not as discontinuing AlphaFold's work, but rather as a shift from a system where dedicated teams solve specific problems to a system where AI, centered around Gemini, supports a wide range of scientific research.