Summary of Key Points
Zhang Yaqin, Dean of the AIR at Tsinghua University, discussed in an interview the future development of AI, strategies for going global, scientific applications, and risk management. She noted that the "ChatGPT moment" for physical AI (such as robots and autonomous vehicles) is still about five years away, and the number of robots could potentially exceed that of humans. Chinese AI companies going global should focus on global challenges like public health and climate change, adhering to the principle of "AI for good," and combining local cooperation with advantages in the industrial chain. The impact of AI on scientific research (such as drug development) is growing exponentially. However, there is also a need to be cautious about potential risks, including technological loss of control, misuse, and systemic social issues, which can only be addressed through multi-stakeholder collaboration for inclusive technology development and safety.
1. The "ChatGPT Moment" for Physical AI: Still 5 Years Away – Will Robots Outnumber Humans?
- What is Physical AI?
Unlike the digital AI we use today (like ChatGPT, which only exists on screens), physical AI can enter the real world in the form of robots, autonomous vehicles, drones, and even AI-powered power grids and logistics systems. It is not intangible; it integrates into our lives, such as household robots that help with cooking or care for the elderly, or self-driving cars on the streets.
- Why Another 5 Years?
Zhang Yaqin believes the "ChatGPT moment" for physical AI will take another five years due to two main reasons:
- Autonomous driving is advancing quickly, but household robots are more challenging: Autonomous driving technology is mature, but what's lacking are large-scale operations and regulatory frameworks (e.g., how to license them and who will be responsible in case of accidents). Household robots face complex environments (identifying different ingredients, handling emergencies with the elderly), requiring foundational models that can handle real-world scenarios, similar to how ChatGPT relies on its large GPT model.
- The scale effect has not yet been reached: In the next decade, everyone may have robots (for industrial, domestic, and social services), potentially exceeding the number of humans. However, these robots are not yet widespread, so we need to wait for the foundational models to mature before a similar explosion in adoption can occur like with ChatGPT.
2. AI Going Global: How Chinese Companies Can Stand Out?
Amidst geopolitical conflicts and trade tensions, Chinese AI companies should not just compete on price but adopt a cooperative and beneficial approach:
- Choose the Right Areas: Focus on solving global issues like public health (AI-assisted disease diagnosis), climate change (AI for optimizing energy use), and aging (AI-powered care robots). These issues are less tied to geopolitics and are more likely to be accepted globally, aligning with the concept of "AI for good."
- Leverage China's Advantages:
- Strong Industrial Chain: The low cost of Chinese AI products is not due to cheap labor but rather high efficiency from automation and robotics.
- Open Source Ecosystem: While large American models (like GPT) are proprietary and expensive, Chinese models like DeepSeek and Qianwen are open source, making them more accessible to small businesses in Africa and other countries.
- Enhanced Innovation: Ten years ago, China was imitating the U.S.; now, many original technologies come from China. By 2025, it is expected that 40% of new drug targets will be developed in China, with countries like the UK and US seeking Chinese expertise.
- Key to Success: Local Cooperation: Going global means adapting products to local needs, such as collaborating with local hospitals for medical AI (addressing data privacy) and hosting models in local data centers (to reduce latency and ensure security), while respecting local laws and cultures.
3. AI's Impact on Scientific Research: Exponential Growth
AI's role in science is transforming; it’s no longer about adding effort but achieving breakthroughs:
- AI Can Work Independently: AI can program, set goals, break tasks down, and iterate through trial and error, significantly accelerating processes (e.g., quickly screening billions of molecules for potential drugs, reducing drug development time from years to months).
- Wide Range of Applications: AI can also verify mathematical formulas, solve physical equations, and assist scientists in experiments. Zhang Yaqin calls this "the greatest technology in human history," with the potential to solve major challenges like healthcare and climate change.
4. Three Major Risks of AI
While AI offers many benefits, there are significant risks:
- Technological Loss of Control: AI algorithms are black boxes, and their decisions can be unpredictable, leading to false information or Deepfakes. If robots or AI systems control critical infrastructure (finance, nuclear power, grids), uncontrolled outcomes could be devastating.
- Misuse of Technology: Malicious actors could use AI for cyberattacks or fraud (e.g., using Deepfakes to impersonate leaders).
- Systemic Social Risks: In the short term, AI may lead to job losses (replacing workers in factories). In the long term, it could change social structures (reducing the need for middle management and requiring reconfiguration of tax systems).
5. Ensuring Access to AI: Overcoming the Digital Divide
AI should benefit everyone, not just a few:
- Open Source Models: Open-source models like Qianwen are affordable in low-income countries, whereas proprietary models from the U.S. are too expensive. African representatives at UN meetings have expressed gratitude for China's open-source ecosystem.
- Exploring New Lifestyles: AI can increase productivity, potentially leading to a Universal Basic Income (UBI), where people work only two or three days a week while earning higher incomes. This requires collaboration among governments, businesses, and scientists to balance innovation and governance.
In summary, the future of AI is promising, but it requires a balanced approach that focuses on both development and safety. We need to integrate physical AI into our lives, help Chinese companies succeed globally, and manage risks to ensure that technology truly benefits all.