On September 30, Zhu Min—former Deputy Managing Director of the IMF and former Deputy Governor of the People’s Bank of China—held an in-depth dialogue with Zhang Yaqin, academician of the Chinese Academy of Engineering, Chair Professor at Tsinghua University, and President of the Institute for AI Industry Research (AIR), focusing on key topics such as the development of AGI and the innovation dynamics between China and the United States. The event was opened by Jiao Jie, Dean of Tsinghua University PBC School of Finance.
Image: Jiao Jie delivering opening remarks
The dialogue began with Zhang Yaqin’s latest book Emergence of Intelligence as the entry point, centering on the core logic of the new wave of AI development. Zhang identified three key concepts driving current AI breakthroughs: Tokenization, Scaling Law, and Emergence Effect. Beyond technological advances, he emphasized the phenomenon of “industrial emergence,” where the “AI+” model is deeply penetrating various sectors. Zhu Min further summarized that this marks a shift in AI’s influence from “curved growth” to “surface diffusion,” enabling comprehensive coverage across diverse industries.
Zhang then forecasted the development timeline for AGI: approximately 4 years for informational intelligence, 10 years for physical intelligence, and 15–20 years for biological intelligence. In response to Zhu Min’s question on human-machine collaboration, Zhang proposed a future vision of AI+HI (Human Intelligence), stressing the importance of human self-awareness and value-driven guidance to steer AI toward benevolence. Both agreed that the evolutionary curve of humans and the rapid progression of AI differ significantly, making their alignment a critical issue for exploration.
On the topic of AI risks and controllability, Zhang offered a rationally optimistic view: while AI algorithms are defined by humans, historical patterns from three industrial revolutions show that despite short-term challenges, the long-term impact of technological innovation is predominantly positive. He expressed strong confidence in AI’s controllability. Zhu Min echoed this sentiment, emphasizing the importance of guiding AI toward public good and inclusivity. He noted that humanity’s ability to follow technology, remain vigilant about risks, and uphold moral integrity form the foundation for addressing future AI-related challenges.
Discussing the competitive landscape between China and the U.S. in AI, Zhang pointed out that since the emergence of models like DeepSeek, the two countries have entered a landscape of differentiated strengths. China has developed distinctive technical paths and business models in low-computing-power scenarios and open-source LLM models. However, he acknowledged that China still lags in computing infrastructure and ecosystem development, though domestic chip companies are accelerating progress through diverse technological approaches. Zhu Min added that China has reached a world-leading level in “AI+” industrial applications and vertical model development. In addition to hardware breakthroughs, initiatives like the “Twenty Measures on Data” and public data openness are strongly supporting China’s AI+ growth. Zhang further analyzed that as the scaling law slows in the pre-training phase, the focus is shifting toward post-training, inference optimization, and agent development—posing new hardware demands and creating more opportunities for China in intelligent agent innovation and scenario-based applications.
Zhu Min noted that while discussions around algorithms and computing power—both horizontal and vertical—have been prevalent, future competition will center on open-source vs. closed-source and business models. Zhang, referencing the mobile internet era, predicted that open-source and closed-source models will coexist long-term. Open-source fosters collective intelligence and accelerates scenario deployment, but in the era of LLM models, it involves multiple layers and differs from the ecosystem logic of the mobile age. On business models, he observed that some U.S. companies have monetized through token mechanisms, while China is more likely to focus on commercializing the AI+ model.
Image: Zhu Min in dialogue with Zhang Yaqin
During the interactive Q&A session, alumni and students actively posed questions. Zhang shared cutting-edge insights on topics such as the limitations of the Turing Test in the multimodal era, interconnection standards for intelligent agents in IoT, and paradigm shifts in AI foundational theory. When asked whether AI possesses fast and accurate intuition, he explained that the human brain has a complex memory system, referencing Daniel Kahneman’s Thinking, Fast and Slow and its concepts of System 1 (fast thinking) and System 2 (slow thinking). He noted that current AI remains largely in the System 2 phase—strong in computing and efficiency, but lacking clear algorithms for intuition, emotion, and consciousness. Zhu Min emphasized that new ideas cannot emerge from nothing; learning remains fundamental, and individuals must learn to build their own cognitive frameworks with the help of LLM models. “Humans must evolve from carriers of knowledge to carriers of thought and wisdom,” Zhu concluded. “In the AI era, the world is undergoing profound transformation, and the demands on individuals are rising. We must trust in our nature, kindness, and capacity to learn.”
This event was co-hosted by Tsinghua University’s PBC School of Finance and the Institute for AI Industry Research (AIR), and organized by the Global Economic Governance 50 Forum.
Image: Dialogue session in progress