Development of Nuclear Science and Technology Enpowered by Artificial Intelligence: Status and Prospects of Artificial Intelligence in China Institute of Atomic Energy
-
YANG Hongyi,
-
LYU Zhaofu,
-
GUO Bing,
-
SONG Qing,
-
ZHU Qingfu,
-
YU Ting,
-
AN Shizhong,
-
REN Lixia,
-
ZHENG Anran,
-
ZHANG Yingxun,
-
GAO Jining,
-
SHANG Chengming,
-
HUANG Peng,
-
WANG Wen,
-
YU Huajin,
-
XIA Yun,
-
WU Xianfeng
-
Graphical Abstract
-
Abstract
This paper aims to systematically review the application status of artificial intelligence (AI) technology across various nuclear energy research fields in the China Institute of Atomic Energy (CIAE), accurately identify existing application gaps, technical bottlenecks, and data barriers, and scientifically formulate CIAE’s future AI R&D roadmap to promote development of nuclear science and technology innovation empowered by AI. Through comprehensive investigation and analysis, AI application practices in CIAE’s core domains, including nuclear physics, nuclear and radiochemistry, reactors, nuclear safety, and nuclear technology applications were focused on. Representative achievements were summarized while deeply analyzing the root causes of insufficient integration between AI and the nuclear science and technology innovation chain/industrial chain. Key constraints such as the shortage of high-quality data, the need for model reliability verification, and deficiencies in computing power and network infrastructure were also evaluated. The research indicates that CIAE has proactively deployed AI applications, establishing super-computing facilities to support digital R&D and achieving breakthrough results in multiple areas. However, significant challenges are as follows: shallow integration between the AI industry chain and the nuclear field; bottlenecks in the scale, quality, and sharing mechanisms of high-quality nuclear domain training data; insufficient explain-ability of existing large models, necessitating extensive rigorous validation of their accuracy, robustness, and safety in nuclear-critical scenarios; gaps in computing power compared to international advanced levels; urgent need for intelligent upgrades of traditional scientific facilities; and inadequate existing network infrastructure to meet future data governance and model training demands. To address challenges and seize opportunities, CIAE will prioritize the following future actions: building a nuclear-intelligence converged innovation system; advancing data governance and construction platform; establishing a nuclear domain foundation model platform; creating a comprehensive nuclear science and technology research platform.
-
-