Summary of Key Points
In just one day, Google's AI division has experienced two significant personnel changes: First, Demis Hassabis, the head of Google DeepMind, shifted from his daily management role to that of a strategic advisor (while still retaining decision-making power over technical matters), and Gemini underwent a leadership transition. Second, Jeff Dean, Google's chief scientist, left the company along with three other core technology experts to start their own business, focusing on AI automation research. These developments caused Alphabet's stock price to drop by nearly 4% in a single day. Although Google has invested in the new company, it also exposes the challenges its AI division is facing, such as delays in model development, talent loss, and the inherent limitations of large corporations.
Detailed Analysis
1. Hassabis' "Retirement from the Front Lines": Not Leaving, but Taking on a New Role
Hassabis has not actually left; he is simply shifting from managing teams and handling day-to-day operations to setting directions and conducting research. His current role includes being the chairman of Google DeepMind and the chief scientist for Alphabet. His main focus areas are:
- Long-term strategy: Developing general artificial intelligence (AGI) that can solve various problems like humans;
- Global AI governance: Collaborating with governments on how to regulate AI to prevent misuse;
- Drug research: He also oversees Isomorphic Labs, a division within DeepMind that uses AI to develop cancer treatments.
The person taking over daily management, Koray, is a familiar colleague who has worked with Hassabis for 13 years and was involved in key projects such as WaveNet (a speech generation technology) and DQN (a classic reinforcement learning project). This adjustment can be seen as an optimization of responsibilities, freeing Hassabis to concentrate on the most critical technological breakthroughs.
2. Jeff Dean's Departure: Taking Away a Major Part of Google's AI Expertise
Jeff Dean is a founding member of Google, having joined the company in 1999 and played a pivotal role in developing its foundational technologies, such as TensorFlow, TPU chips, and distributed storage systems. The three experts he brought with him are also top-tier professionals:
- Sanjay: Co-created Google's file system;
- Oriol: Responsible for Gemini's technology and a key member of AlphaFold, which uses AI to predict protein structures;
- Quoc: Part of Google Brain and involved in AutoML projects that aim to make AI develop its own algorithms.
Together, these four individuals cover all critical aspects of Google's AI efforts. Their departure represents a significant loss of technical expertise, which is why the market reacted so strongly.
3. The New Company Discovery Loop: Automating Scientific Research with AI
The new company, Discovery Loop, aims to automate the entire scientific research process. Currently, scientists have to propose hypotheses, design experiments, and analyze results manually, which can be time-consuming (days to months). Discovery Loop intends to use AI to:
- Automatically generate research hypotheses;
- Conduct large-scale calculations for experiments (such as drug molecule screening and chip design);
- Analyze results and iterate on the experimental process.
This approach could significantly speed up research, potentially reducing the time required to develop new drugs from ten years to just a few years or even less. This direction targets an unexplored area in AI and holds great potential.
4. Why Google Invested in the Departing Employees' Company
Despite the departure of key employees, Google chose to invest in their new company and establish a cloud service partnership. The reasons are practical:
- Maintaining access to their expertise: These individuals are industry leaders, and investing in them allows for continued technical collaboration (e.g., using Google's cloud services). There is also the possibility of acquiring their research findings in the future;
- Investing in a promising trend: AI automation in research is a future direction. Google itself is working on drug research through Isomorphic Labs, so investing in this company aligns with its strategic goals;
- Stabilizing the team: By publicly supporting the departing employees, Google aims to reduce anxiety among remaining staff and demonstrate its commitment to retaining top talent.
This move is both strategic and graceful, minimizing the impact of the talent loss.
5. Challenges Faced by Google's AI Division
These changes reflect the recent difficulties faced by Google's AI division:
- Slow model development: Gemini 3.5 Pro was scheduled for release in June but has yet to appear, leading to growing market skepticism about Google's progress (the stock price dropped 7% two weeks ago due to increased AI-related expenses without clear results);
- Severe talent loss: Previous key contributors have joined other companies (Noam to OpenAI, John to Anthropic), indicating that Google struggles to retain top talents;
- Limitations of large corporations: With its vast workforce and complex processes, Google's infrastructure is not well-suited for rapid AI iteration. The new company's founders pointed out that Google's infrastructure is designed for stable services (such as search and advertising) rather than experimental research.
Google needs to either adjust its internal mechanisms to retain talent and accelerate model development or risk losing its leading position in the AI field as more experts leave.
In Conclusion
Google's AI division is at a crossroads. It must either improve its internal processes to retain top talents and accelerate model development or watch as more experts move on, potentially leading to a gradual loss of its leadership in the AI industry. These personnel changes serve as both a warning and an opportunity for Google to address these challenges.