吴仲琦,刘胤杉,代涛,郑晓龙.大模型关键技术攻关与产业化发展的启示与建议[J].中国科学院院刊,2026,41(8):1683-1692.

大模型关键技术攻关与产业化发展的启示与建议

Critical core technology breakthroughs in large-scale models: Industrialization strategies and policy implications

吴仲琦^1,2
中国科学院科技战略咨询研究院 北京 100190;北京工商大学 计算机与人工智能学院 北京 100048
WU Zhongqi^1,2
Institutes of Science and Development, Chinese Academy of Sciences, Beijing 100190, China;School of Computer and Artificial Intelligence, Beijing Technology and Business University, Beijing 100048, China
刘胤杉¹
中国科学院科技战略咨询研究院 北京 100190
LIU Yinshan¹
Institutes of Science and Development, Chinese Academy of Sciences, Beijing 100190, China
代涛^1,3
中国科学院科技战略咨询研究院 北京 100190;中国科学院大学 公共政策与管理学院 北京 100049
DAI Tao^1,3
Institutes of Science and Development, Chinese Academy of Sciences, Beijing 100190, China;School of Public Policy and Management, University of Chinese Academy of Sciences, Beijing 100049, China
郑晓龙^4,5*
中国科学院自动化研究所 多模态人工智能系统全国重点实验室 北京 100190;中国科学院大学 前沿交叉科学学院 北京 101408
ZHENG Xiaolong^4,5*
State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China;School of Advanced Interdisciplinary Sciences, University of Chinese Academy of Sciences, Beijing 101408, China

         大模型;关键核心技术攻关;产业生态;技术治理

        large-scale models;key core technology breakthroughs;industrial ecosystem;technology governance

        大模型技术作为人工智能领域的关键技术突破方向，其攻关效能关乎国家科技战略主动权的得失。文章基于“技术突破—产业转化—治理政策”多维度框架，系统揭示大模型技术的突破路径与产业化瓶颈：在技术层面，大模型在参数规模与算力需求上呈现指数级增长，但面临高质量数据枯竭、迁移学习能力不足、底层理论创新不足等瓶颈问题；在产业层面，大模型正重塑全球产业链格局，推动“通用化+专业化”双轨应用，但生态碎片化、人才短缺与安全风险制约其规模化落地；在政策层面，国际竞争加剧与技术治理复杂性凸显，亟须构建动态政策框架与全球协作机制。针对上述挑战，文章提出强化算力与算法自主创新、深化产学研协同、健全风险治理体系等系统性建议，以推动大模型技术的高质量发展，为战略性产业升级提供支撑。

        As a pivotal direction for breakthroughs in key core technologies within the artificial intelligence domain, large-scale models hold strategic significance in securing national scientific and technological sovereignty. This study employs a multidimensional framework encompassing “technological breakthroughs, industrial transformation, and governance policies” to systematically investigate the developmental trajectories and industrialization bottlenecks of large-scale models. At the technological level, while large-scale models exhibit exponential growth in parameter scale and computing power demands, they face critical challenges including the scarcity of high-quality data, insufficient transfer learning capabilities, and reliability-explainability trade-offs. Industrially, these models are reshaping the global industrial chain landscape through a dual-track application pattern integrating general-purpose and specialized models, yet their scalable deployment is constrained by a fragmented ecosystem, talent shortages, and security risks. From a policy perspective, intensified international competition and governance complexities necessitate the establishment of dynamic policy frameworks and global collaborative mechanisms. To address these challenges, this study proposes systemic strategies to strengthen indigenous innovation in computing power and algorithms, deepen industry-academia-research collaboration, and improve risk governance frameworks. These recommendations aim to foster the high-quality development of large-scale models and provide strategic support for industrial upgrading.