美军人工智能系统试验鉴定框架分析与启示  

Analysis and enlightenment of U.S.military artificial intelligence system test and evaluation framework

作　　者：刘宏亮[1] 沈卫 LIU Hongliang;SHEN Wei(North Science and Technology Information Institute,Beijing 100089,China)

机构地区：[1]北方科技信息研究所,北京100089

出　　处：《国防科技》2025年第1期80-88,共9页National Defense Technology

摘　　要：阐述美军人工智能系统试验鉴定的顶层战略及其衍生的行动路径。以人工智能系统全生命周期流程为时间轴,模型试验鉴定、系统集成试验鉴定、人机集成试验鉴定和作战试验鉴定为维度轴,阐述美军人工智能系统试验鉴定框架主要内容,从中归纳美军人工智能系统试验鉴定的3点启示:重视数据的全流程驱动作用;通过试验鉴定“左移”降低系统集成风险;通过试验鉴定“右移”保障人工智能系统性能。The top-level strategy and the resultant action path of the U.S.military artificial intelligence system test and evaluation are delineated in this paper.With the entire life cycle process of the artificial intelligence system serving as the time axis and model test and evaluation,system integration test and evaluation,human-system integration test and evaluation,and operational test and evaluation as the dimensional axis,the principal contents of the test and evaluation framework of the U.S.military artificial intelligence system are analyzed,and three inspirations from the test and evaluation of the U.S.military artificial intelligence system are summarized:Pay heed to the full-process driving role of data,mitigate the risk of system integration through the test and evaluation“left shift”,and ensure the performance of the artificial intelligence system through the test and evaluation“right shift”.

关 键 词：美军 人工智能 试验鉴定框架 集成试验 作战试验鉴定 

分 类 号：N945[自然科学总论—系统科学] TP18[自动化与计算机技术—控制理论与控制工程]