Committee Channel | Zhou Zhihua: Leading Paradigm Shift in Scientific Research with Artificial Intelligence

The Fourth Session of the 14th National Committee of the Chinese People’s Political Consultative Conference held its second plenary session on the afternoon of the 7th, with members delivering speeches at the conference. Xinhua News Agency provided online text and image live coverage.

Zhou Zhihua, Vice President of Nanjing University and Academician of the Chinese Academy of Sciences:

Dear members, I would like to speak on behalf of non-party members on the topic: Leading Paradigm Shift in Scientific Research through Artificial Intelligence.

“Empowering scientific research with artificial intelligence” is driving a historic transformation in research paradigms. It is regarded as the “fifth scientific research paradigm” after experience, theory, computation, and data. This not only accelerates solving long-standing major scientific problems but also has the potential to reshape the fundamental pathways of scientific discovery and significantly improve original innovation efficiency. In August 2025, the State Council issued the “Opinions on Deepening the Implementation of the ‘AI+’ Action,” explicitly listing “AI+” science and technology as one of the key focus areas. Currently, some scientific research efforts remain at the stage of simple tool application or blindly attempting to train general “large scientific models” to address all problems. Meanwhile, high costs of data acquisition, inconsistent standards, low willingness to share data, uneven data annotation quality, and the lack of authoritative, standardized, large-scale scientific datasets lead to low training efficiency and reliability of AI models, with prominent issues of redundant development and resource waste.

Therefore, I propose:

1. Strengthen policy guidance and enhance basic innovation capabilities. Optimize the overall layout of scientific research in the AI field, avoid excessive resource concentration on compute-intensive applications, correct the misconception of “solving everything with large models,” and increase support for fundamental AI algorithm research. Improve the innovation ability of designing solutions for specific problems. Focus on supporting a batch of forward-looking, strategic basic research projects, and encourage researchers to conduct original research. At the same time, guide enterprises and social capital to participate in fundamental AI algorithm research, forming diversified investment mechanisms. Establish a scientific and reasonable scientific evaluation system and create an environment that encourages exploration and tolerates failure.

2. Transform training models and build a multidisciplinary talent team. From the source, establish a “dual-degree” training system for innovative talents empowered by AI, supporting pilot programs at high-level research universities to set up “PhD + Master’s” dual degrees. Support PhD students pursuing a scientific master’s degree in a different field during their AI doctoral studies, exploring new interdisciplinary graduate training models, and systematically cultivating “bilingual” scientists who are proficient in their domain and cutting-edge AI technologies. Additionally, create “interdisciplinary special zones” in degree awarding, professional title promotion, and performance evaluation to address the “two ends not relying” dilemma faced by interdisciplinary talents under traditional evaluation systems.

3. Focus on two-way science popularization to eliminate disciplinary barriers. Build an interdisciplinary “two-way translation” and collaboration mechanism. On one hand, scientists in the field translate scientific problems into language understandable by AI researchers, expressing key frontier issues for targeted solutions. On the other hand, AI scientists conduct science popularization for traditional domain scholars, clarifying technical boundaries through concrete cases, alleviating fears or blind worship of AI, and explaining AI as a tool to assist scientific discovery. Encourage regular cross-disciplinary salons to promote交流 among scholars from different backgrounds and facilitate a transition from concept dissemination to substantive collaboration.

4. Strengthen data governance and build a scientific data ecosystem. Led by relevant national departments, establish a national-level scientific data sharing and service platform. Rely on major science and technology platforms such as key laboratories to build standardized scientific data warehouses, formulate standards for data collection, annotation, storage, and sharing across disciplines, and introduce data quality feedback mechanisms to continuously optimize data assets. Through project funding and成果评价 policies, encourage research institutions and researchers to open and share scientific data, forming a healthy ecosystem and maximizing utilization efficiency. At the same time, enhance support for技术研发 and legal regulations to effectively protect sensitive information and intellectual property during data sharing.

(Edited by: Wen Jing)

Keywords: