Summary of Key Points
The core AI research team at DeepMind, which won the Nobel Prize in Chemistry in 2024 for its development of AlphaFold, was disbanded in less than two years. Team members either moved on to work on the general-purpose large model Gemini or joined Isomorphic Labs, an AI pharmaceutical company under Alphabet, with nearly a quarter of them leaving the company altogether. This is not a case of DeepMind “breaking bridges after crossing the river” but an inevitable outcome of the competitive landscape in the AI industry entering a resource-intensive phase. Large companies are prioritizing their most valuable talents, computing power, and funding on general-purpose large models that can quickly respond to market competition, while delegating the practical implementation of AI research to more commercialized teams.
Detailed Explanation
1. AlphaFold: The Scientific Revolution
AlphaFold was no ordinary AI tool; it solved a long-standing problem in biology—the folding of proteins. Proteins are the “building blocks” of life, but their functions are determined by their three-dimensional structure, not just the sequence of amino acids that make them up. Previously, scientists spent months or even years using expensive methods like cryo-electron microscopy and X-rays to determine protein structures. AlphaFold used AI to predict protein structures directly from amino acid sequences, achieving near-experimental accuracy for the first time in 2020. In 2021, its database was made public (now containing over 200 million protein structures), fundamentally changing the approach to structural biology research. Scientists can now focus on identifying disease causes and developing drugs instead of spending time on experiments. AlphaFold’s Nobel Prize recognition marked a shift from AI being a tool for efficiency improvement to a driving force for scientific discovery, embodying DeepMind’s commitment to long-term projects with uncertain immediate benefits.
2. Team Disbandment: Not Abandonment, but Resource Reallocation
The main reason for DeepMind’s dissolution of the AI4S team was the intense competition from general-purpose large models. The current focus of the AI industry is on these models (such as OpenAI’s GPT and Anthropic’s Claude), with Google’s Gemini being a direct competitor. However, Gemini’s progress has been slow, leading to concerns that Google might fall behind. As a result, DeepMind had to reallocate its top talents and resources to Gemini.
3. Reorientation of Research
DeepMind did not abandon AlphaFold; instead, it handed over its development to the more commercialized Isomorphic Labs. Founded in 2021, Isomorphic specializes in AI-driven drug research. Its goal is to apply AlphaFold technology to create practical commercial value, collaborating with pharmaceutical companies like Novartis and Eli Lilly to develop drugs for difficult targets. For example, while AlphaFold can predict protein structures, Isomorphic aims to design small molecules that can bind to these proteins, potentially accelerating drug development. Although no AI-driven drugs have yet been released, this represents a critical step in moving AI research from the lab to practical applications.
4. Industry Trends
DeepMind’s decision is representative of the broader trends in the AI industry. More tech companies are reevaluating their AI research strategies. Training large models requires massive computing power (GPUs), top-tier talent, and continuous funding. Companies must determine which projects can quickly generate market share (like general-purpose models) and which require long-term investment with no immediate returns (such as basic AI research). As a result, they are shifting the focus of basic research to more specialized teams while concentrating on developing competitive large models.
5. Future Directions
The dissolution of the core AlphaFold team means that breakthroughs like AlphaFold 3 may not occur in the same way, but AlphaFold’s technology will likely have a broader impact on everyday life. If Isomorphic Labs succeeds in using AI to develop new drugs, we could see faster market launches for treatments for cancer and rare diseases at lower costs. DeepMind’s focus on Gemini will also enhance the capabilities of general-purpose models, making AI more useful in scientific research, coding, and daily tasks. This transition is essential for AI to move from a laboratory-based technology to a widely applied solution.
Conclusion
The dissolution of the AlphaFold team marks a shift in AI research from demonstrating its capabilities to creating tangible value. Large companies are focusing their resources on general-purpose models, while the results of AI research will be realized through commercial applications. While we may no longer hear about future versions of AlphaFold, we are likely to see AI-driven technologies in drugs and tools that improve our lives.