The rapid development and maturation of Artificial Intelligence (AI) technology in recent years have led to a surge in the adoption of AI in military hardware. This has considerably expanded the capabilities of conventional military hardware, especially in terms of autonomy and intelligence. However, new challenges have also been created in terms of quality supervision, as AI algorithms often exhibit intrinsic flaws in their logical reasoning, lack explainability, and require training. The weaknesses of current assessment methods have resulted in many practicality and reliability issues in the development of AI hardware. By reviewing the novel characteristics of AI hardware and the new challenges they create for quality supervision, this study discusses how the quality supervision of AI hardware may be conducted in an effective manner. Six suggestions are provided for improving quality supervision for AI hardware: strengthen basic capacity and infrastructure, increase attention to daily data collection and collation, implement flexible technical-status management methods, develop effective assessment methods, improve talent management, and establish information feedback loops. These suggestions may serve as a reference for the development of AI hardware.
张礼学,杨 勇,周 琳,蔡志刚.【专题】人工智能装备质量监督效能研究[J].国防科技,2021,42(1):123-127；ZHANG Lixue, YANG Yong, ZHOU Lin, CAI Zhigang. Improving the efficiency of quality supervision for Artificial Intelligence hardware[J]. National Defense Technology,2021,42(1):123-127复制