Nearly 50 academicians and more than 80 young scholars gathered in Beijing on August 21-22 for the 2026 AI for Science Conference, a two-day event that drew 4,000 attendees to the Zhongguancun International Innovation Center. Themed "AI Transforming Research Paradigms," the conference used 2 plenary sessions and 14 parallel sessions to examine how artificial intelligence is reshaping the practice of scientific research across fundamental science, industry applications, and international collaboration.
AI4S enters its "second half"
Weinan E, Chair Professor at Peking University's School of Mathematical Sciences and a CAS academician, opened the first plenary session with a keynote titled "Science after AI for Science." He argued that the field has moved past the exploration phase and into systematic transformation of the research enterprise itself.
E took direct aim at the current incentive structure: the research system, with publication as its core metric, "has objectively diluted the original intent of 'free exploration' and 'scientific discovery.'" He called for democratizing scientific research, connecting talent, innovation, markets, and capital from an industrial perspective, and establishing low-cost, high-efficiency research tools that let capable individuals and teams participate in scientific innovation.
Other first-day speakers traced the full-stack AI4S picture. Zhang Linfeng, Founder and Chief Scientist of DP Technology, said implementation requires infrastructure first - foundational scientific large models, high-quality scientific data, and autonomous laboratories. Wang Jian, Director of Zhejiang Lab and founder of Alibaba Cloud, argued that tokenized scientific data is crucial for AI for Science, and that multidisciplinary barriers "will be crossed by more researchers." Xie Xiaoliang, Director of Changping Laboratory, analyzed the post-AlphaFold era in biomedicine, where precise biomedical big data is the key input for AI-driven discovery.
From lab to industry
The second day's plenary focused on how AI4S can deliver concrete results. Speakers covered materials large models, AI-driven autonomous laboratories, and oil industry implementation. Wang Haifeng, Chief Technology Officer of Baidu, offered an industry perspective on deep learning and scientific intelligence.
The 14 parallel sessions ran across three tracks. The Fundamental Research track covered AI4S foundational theory, scientific data, research agents, and autonomous laboratories. The Integration of Technological Innovation and Industrial Innovation track covered biomedicine, energy and chemicals, materials development, scientific instruments, and industrial intelligence - including intelligent drug development, intelligent chemical laboratories, and unmanned factories.
The International Scientific and Technological Cooperation track anchored its sessions on four major science programs led by Chinese scientists: Panoramic Digital Life, Digital Sustainable Planet, PLANeT, and Deep-time Digital Earth.
Three announcements mark policy progress
The conference served as the release platform for several policy and standards milestones. Beijing released its first batch of typical AI-empowered scientific research cases, following the city's 2026-2028 implementation plan for accelerating AI-empowered research. The collection spans autonomous laboratory systems, high-value benchmark applications, scientific model systems, scientific data foundations, and innovation ecosystem development.
Haidian District launched the Zhongguancun (Haidian) AI4S Innovation Cluster in the Xisanqi area, coordinating 1.6 million square meters of industrial space and 740,000 square meters of pilot-scale testing space. The district plans to use the cluster to develop specialized chips for scientific intelligence, build autonomous laboratory systems, create AI-native instruments, and establish a dedicated AI4S fund.
Li Shunchao, Director of the Information Technology Division of the Beijing Municipal Science and Technology Commission, reported progress on China's first national standard for autonomous laboratories. The standard uses a four-layer framework: device abstraction, intelligent management, intelligent experimentation, and safety control.
Why this matters for science and research professionals
For researchers, the conference's core message is that AI4S is moving from pilot projects to institutional infrastructure. The release of Beijing's first typical cases and the launch of the Haidian innovation cluster signal that government funding and policy support are now tied to concrete implementation targets, not exploratory grants. The autonomous laboratory standard, once finalized, will define how labs report data and integrate AI tools - which means researchers who align their workflows with these standards early will have an advantage in future funding cycles. For those tracking the field, the shift in emphasis from methodology papers to platform-based research and "wet-dry closed loops" suggests that the next round of major scientific results may come from teams that combine domain expertise with AI infrastructure, not from either alone. Professionals looking to build these skills can explore AI for Science & Research resources and AI Research Courses to understand how these tools apply to their own work.
Your membership also unlocks:
