WAIC holds AI for science forum

Leading scientists and AI pioneers from around the world convened in Shanghai on Saturday for an AI-for-science-themed forum at the 2026 World AI Conference, mapping out a vision for integrating artificial intelligence into scientific research.

Under the theme "AI-native closed-loop discovery", the forum brought together Nobel laureates, university leaders, and technology innovators to address how scientific institutions should evolve as AI becomes increasingly central to research methodology.

"Scientific intelligence cannot remain merely a tool — it must become an ecosystem," Peng Huisheng, professor at Fudan University and academician of Chinese Academy of Sciences, said. "We need to build advanced discovery systems where humans and machines collaborate, where data and experiments interconnect, and where problems and validations form closed loops."

Peng emphasized that discovery comes not from isolated breakthroughs, but through integrated platform capabilities. "Talent cultivation isn't about teaching students to use AI — it's about cultivating people who can cocreate with AI," he explained.

Wang Jian, academician of the Chinese Academy of Engineering, director of Zhejiang Lab, and founder of Alibaba Cloud, echoed this vision. "Technology should augment human capabilities, not simply replace them," he said.

Wang predicted that AI is becoming as foundational as mathematics itself, driving a paradigm shift from STEM to "STE+MAP" — science, technology, engineering plus math, AI, and public facilities. "Scientific intelligence lies not in paper texts, but in data," he said. "Once we convert diverse scientific data into unified vectors, disciplinary boundaries will dissolve. AI could ultimately eliminate the concept of interdisciplinary research itself, making science whole again."

Nobel laureate Arieh Warshel demonstrated how AI accelerates scientific discovery when paired with theoretical understanding. Drawing from his work on enzyme catalysis, he described how his team has applied hybrid quantum-mechanical and molecular-mechanical modeling approaches, enhanced by maximum entropy analysis, to guide enzyme engineering and protein design.

"AI has great potential for industrial design of these systems," Warshel said, while cautioning that effective design requires combining AI tools with deep physical and chemical knowledge.

Fan Jianqing, professor at Princeton University and member of the US National Academy of Sciences, described AI as a dynamic cycle of statistical learning and optimal decision-making, illustrating how AI transforms data into social insights.

"The real challenge is not making machines smarter, but ensuring intelligence — whether human or artificial — serves society," Fan said. "AI is not only making machines think, but also helping society reveal unknowns in our world and improve social well-being."

He stressed that advancing intelligent agents and scientific innovation must go hand-in-hand with addressing problems such as employment disruption, educational transformation, ethical safety, and maintaining effective human control.

Gilles Brassard, professor at the Université de Montréal and 2025 AM Turing Award winner, discussed quantum computing's impact on traditional cryptography and data security, and expressed high expectations for its potential in emerging fields.

When offering advice to young scientists, Brassard urged them to follow their curiosity. "Choose research topics you find interesting, not because they seem practical," he said. "Einstein and the quantum physics pioneers had no idea their theories would transform society. Imagination is more important than knowledge."

A panel discussion on original innovation in the AI-native era examined scientific intelligence through multiple lenses. Participants agreed that original innovation in the AI-native era depends on whether people, data, models, experiments, and academic judgment can form a novel collaborative discovery mechanism.

The forum unveiled the Fuxi weather constellation technology initiative, which leverages Fudan's intelligent weather forecasting model and a newly built AI-meteorology data center. The system has accumulated over a decade of AI-ready datasets to complete the chain from observation to forecasting on-orbit.

The forum also marked the inaugural publication of two academic journals — Science and AI and AI for Engineering — establishing new platforms for academic exchange spanning from fundamental discovery to engineering applications in scientific intelligence.