摘要
人工智能驱动的科学研究(AI for Science)被视为科学发现的第五范式的曙光。依循演绎主义的科学研究逻辑,梳理了人工智能在科学假设生成、数据收集以及分析挖掘中的应用。人工智能“数据算法算力”三原则,对科学数据的质量、算法的复杂性以及计算能力提出了更高的要求。AI for Science时代预计会出现科技巨头、AI专家、软硬件工程师、政府以及教育机构等紧密协同的新型科研模式。然而,AI算法的黑箱特性对科学研究的可解释性和可重复性构成潜在威胁。因此,在推进人工智能驱动的科学研究的发展过程中,必须坚持伦理优先的原则,注重科学数据的安全性管理,防范化解大模型分布外泛化带来的解释性弱等问题。
Artificial Intelligence(AI)driven scientific research,known as AI for Science,is heralded as the dawn of the fifth paradigm in scientific discovery.Following the deductive scientific research logic,this paper outlines the applications of AI in the generation of scientific hypotheses,data collection,and the analysis and mining of data.Under the triad of AI principles—“data,algorithms,and computing power”—higher standards are set for the quality of scientific data,the complexity of algorithms,and computational capabilities.The AI for Science era is anticipated to foster a new type of collaborative research model,involving tech giants,AI experts,software and hardware engineers,governments,and educational institutions in close coordination.However,the black-box nature of AI algorithms poses a potential threat to the interpretability and reproducibility of scientific research.Therefore,in advancing the development of AI-driven scientific research,it is imperative to uphold the principle of ethics first,focus on the secure management of scientific data,and guard against and mitigate issues such as the weakened interpretability caused by the out-of-distribution generalization of large models.
作者
王晨阳
褚建勋
WANG Chenyang;CHU Jianxun(School of Humanities and Social Sciences,University of Science and Technology,Hefei 230051,China;Institute of Computational Social Science and Media Studies,University of Science and Technology,Hefei 230051,China)
出处
《南京邮电大学学报(社会科学版)》
2024年第4期10-19,共10页
Journal of Nanjing University of Posts and Telecommunications(Social Science Edition)
基金
中国科协研究生科普能力提升项目“伦理风险型技术的科普解读与社会感知研究——以人工智能医疗技术为例”(KXYJS2022070)
中国科协研究生科普能力提升项目“乡村振兴视角下基层科技社团参与公共服务模式与对策研究”(KXYJS2022068)。