Google DeepMind Disbands AlphaFold Team to Shift Focus to Gemini
Google DeepMind has disbanded the original AlphaFold team, with most authors of the past year's papers reassigned to projects such as Gemini, enzyme design, nuclear fusion, and genomics. Nearly a quarter of the full-time authors have left the company.
AlphaFold can predict protein 3D structures, helping Demis Hassabis and John Jumper win the 2024 Nobel Prize in Chemistry. It was previously a representative of DeepMind's concentrated team tackling a single scientific problem; now, the focus is shifting towards creating a general scientific assistant with Gemini to automate parts of the research process.
In June this year, Jumper announced his move to Anthropic, and core members Jonas Adler and Alexander Pritzel will also join him. The results of AlphaFold will still be used by Isomorphic Labs for drug development. Under this event-driven scenario, top scientific talent is concentrating on general large models and competitors, benefiting the Gemini ecosystem and Anthropic, while putting pressure on the specialized scientific problem-solving model.
Source: Public Information
ABAB AI Insight
DeepMind achieved breakthroughs in protein folding with AlphaFold and won a Nobel Prize, historically focusing on a single team tackling vertical scientific problems. Now, through personnel reallocation, it is shifting towards a Gemini-driven general scientific assistant, reflecting a strategic adjustment from specialized breakthroughs to platform-based automation.
In terms of capital, resources are being internally allocated to enzyme design, nuclear fusion, genomics, and Isomorphic Labs, while core talent is flowing to Anthropic. The motivation is to concentrate resources to create large models that can assist multi-domain research, reducing reliance on single projects and accelerating the automation of research processes.
Similar cases can be seen with OpenAI and Anthropic transitioning from specialized models to general agents. Currently, AI research is in a phase of expanding from vertical tools to general assistants, with talent competition becoming a key variable.
This essentially represents a technological replacement: general large models are replacing specialized teams tackling problems, with the mechanism being embedding domain knowledge into scalable platforms, shifting scientific discovery from labor-intensive to model-assisted scalable processes.
ABAB News · Cognitive Laws
Nobel Prize teams will also be reconfigured
General models are consuming specialized breakthroughs
Talent flow determines research focus