Summary of Key Points
In the era of AI, top scientists such as Jeff Dean from Google and Yann LeCun from Meta have left large companies to start their own businesses. The root cause is the increasing conflict between personal research goals (freedom to explore, verification of unique technical approaches, pursuit of long-term value) and corporate constraints (pressure to deliver products, strategic alignment requirements, limited resource allocation). Through an interview with Silicon Valley HR expert Tom Zhang, this article compares the talent management philosophies of Google, Tesla, and Tencent, discussing issues such as how non-expert managers can manage professionals, evaluate the value of scientific research, and the challenges Chinese companies face in identifying talent. It provides insights for businesses to address the talent challenges of the AI era.
Why Do Top Scientists Want to Leave Large Companies? — The Conflict Between Personal Aspirations and Corporate Constraints
Top scientists leave not because of low salaries or poor job prospects, but due to fundamental differences in values:
- What do they want? They have their own “north stars”—for example, Yann LeCun aims to develop world models that can understand the physical world, while others focus on foundational large models and embodied brains. These are cutting-edge areas that require long-term investment with no immediate commercial returns.
- What do companies need? Large companies are under pressure to meet KPIs: products must be released on time, investments must yield results, and research directions must align with corporate strategies (for instance, Meta may prioritize quickly monetizable large language models over LeCun’s world models).
- The AI era exacerbates this conflict: In the past, large companies could attract scientists with stable funding and ample computing power. However, with rapid technological advancements, research outcomes are now directly tied to product competitiveness and capital investment, making the personal influence of scientists much more significant than their job titles.
Talent Management at Google, Tesla, and Tencent: Three Completely Different Approaches
All three companies compete for top talent, but their management styles differ dramatically:
- Google: Empowering Intelligence
The core philosophy is that intelligent individuals do not need to be micromanaged. Managers use data to persuade (e.g., the Project Oxygen study on the traits of excellent managers), and major decisions are made through collective discussion. The approach emphasizes giving employees autonomy, funding, and a sense of security, allowing them to pursue their research interests.
- Tesla: Mission-Driven with High Pressure
Elon Musk sets goals (e.g., “going to Mars”), and the team focuses on finding the best ways to achieve them. The organization is highly hierarchical, with Musk directly communicating with frontline engineers to drive breakthroughs (code must be ready for use, products must be viable). Only those willing to dedicate themselves to the mission stay; others are eliminated.
- Tencent: A Practical Dual Track System
There are separate paths for technical expertise (T-channel) and management (M-channel). You can focus on technology or manage others. The company demands both rapid product delivery and room for experimentation (e.g., with AI research projects). The hiring criteria are based on proven capabilities.
How Can Non-Experts Manage Researchers? — Set Clear Goals First
How can non-technical managers oversee scientists effectively? The key is to focus on the big picture:
- Set clear goals: Just as Kennedy aimed for a moon landing within 10 years, Musk aims for Mars. Managers don’t need to understand the details; they just need to communicate the vision and let experts handle the specifics. For example, when Liang Wenfeng at DeepSeek set the goal of developing AGI, the team worked towards that goal.
- Manage goals, resources, and boundaries, not the details: Managers should focus on whether a project aligns with corporate strategy and what resources are needed (e.g., how many GPUs) and whether there are any security risks. If they don’t understand technology, they should consult experts.
- Act as Investors but with Limits: When scientists request resources (e.g., 10,000 GPUs), managers should assess the feasibility rather than cutting costs arbitrarily. They also need to establish ethical and safety guidelines (e.g., avoiding the use of unauthorized data).
How to Evaluate the Value of Scientific Research? — Beyond Papers and Patents
Companies should evaluate research beyond the number of papers or patents; focus on practical impact:
- Don’t Overemphasize Quantity: IBM led in patents for 29 years but failed to translate that into significant product success. Some teams publish many papers without producing useful products, which is a form of “formalism.”
- Focus on Impact: Projects like AlphaGo and AlphaFold have made real contributions (changing the game of Go, solving protein structure problems). Even if a project fails, the new directions discovered along the way (e.g., the invention of sticky notes) should be recognized as valuable.
- Allow for Mistakes but Don’t Indulge in Neglect: Research involves uncertainties; failures are acceptable as long as they are reasonable. Google provides more time for cutting-edge research, while Tesla prioritizes quick validation for short-term tasks. The focus is on progress and credibility.
Challenges for Chinese Companies in Identifying Talent: **Don’t Rely Solely on “Big Company Labels”**
Chinese companies often seek talented individuals with a “big company aura,” but this can lead to mistakes:
- Overreliance on Labels: They focus on whether someone comes from Google or OpenAI or has an overseas background, neglecting their actual capabilities and fit within the team. DeepSeek, for example, hired local talents who achieved results without the label of a big company.
- Focus on Performance: Talent is proven through action, not just interviews. Huawei emphasizes that leaders are developed through practical tasks, not just resumes.
- What About Small Companies? Start by clarifying business goals, use existing talent effectively, and build success before attracting better candidates. Don’t try to assemble a dream team from the start; build a solid foundation first.
Conclusion
In the AI era, the main reason large companies cannot retain top scientists is the mismatch between their goals. To solve this issue, companies need to adopt the right management styles (e.g., Google’s freedom or Tesla’s mission-driven approach), teach managers to focus on the big picture, establish fair evaluation systems, and let go of the obsession with “big company labels.” After all, talent is not scarce; what’s truly scarce are the skills to identify and utilize it effectively.