WAIC Day 1: New Insights on AGI and Agents

量子位 · wechat · 2026-07-18

Key Signals from Multiple Guests on Day 1 of WAIC

John Hopcroft believes that in the AI era, a nation's true core competitiveness is talent; the fundamental mission of universities should be cultivating the next generation of talent, not just conducting research.
Yin Qi predicts that the next AI industry boom will come from the combination of agent models + next-gen terminals. Agents will become the smallest unit of productivity, driving the industry into its "third wave" of evolution.
Zhou Bowen proposed that Research is the next Coding; scientific research is not just an application scenario but likely the ultimate challenge for AGI. Shanghai AI Lab also released the Intern·Duanyan Scientific Discovery Platform.

Perspectives on Model and Intelligence Routes

Richard Sutton believes that static data labeling is nearing its ceiling, and the core route for next-gen AI is experience-driven—meaning agents learn continuously through interaction with their environment.
Su Hao emphasized that the key to solving LLM hallucinations lies in physical intelligence: letting models enter the real world to accumulate verifiable experience through prediction, action, and feedback correction.
Wang Jian pointed out from a scientific infrastructure perspective that scientific data could be the key driver of the next AI revolution, and foundational science models need to integrate data, papers, and code.

Safety and Governance

Xu Li believes that the core value of AI is amplifying individual capabilities rather than replacing human labor, and future billing models might shift from Token to Task.
Bengio stressed that both the utility and the risks of agents stem from their autonomy; high-risk AI should adhere to "proving safety before allowing deployment," and governance cannot rely solely on corporate self-regulation.

More from AGI Musings