摘要
生成式人工智能作为通向通用人工智能的核心驱动力,通过深度模拟人类认知中的创造性思维与自适应学习机制,为构建具备多模态内容生成、复杂问题求解能力的智能系统奠定了范式基础。对2025年生成式人工智能的主要发展动向进行综合评述。首先,梳理生成式人工智能技术的发展现状;其次,概述生成式人工智能技术的军事应用;最后,阐释对该技术发展的相关思考。综述表明,多模态大模型训练与推理技术、3D人物和场景自动生成技术、推理计算加速与低资源硬件适配技术,以及生成式具身环境感知与交互技术,是当前该领域的四大研究热点。实证表明,该技术在战场态势动态感知、多源情报融合整编、智能决策支持等军事场景中展现出显著效能。据此,亟需从顶层战略规划、技术能力供给、基础支撑体系三位一体推进,构建面向智能化应用需求的技术发展新范式,以实现生成式人工智能在军事领域的战略性突破与可持续创新。
Generative Artificial Intelligence(Generative AI),as the core driving force towards Artificial General Intelligence(AGI),lays a paradigm foundation for constructing intelligent systems with multi-modal content generation and complex problem-solving capabilities,by deeply simulating creative thinking and adaptive learning mechanisms in human cognition.Specifically,this work reviews some key development trends of Generative AI in 2025.Firstly,this paper analyzes the development trend of GAI,building on this analysis,it then systematically summarizes the applications of the technology in the military fields.Finally,it explores the relevant thoughts regarding the development of this technology.Based on this structure,this survey identifies that current research in this field focuses on four core topics:multi-modal large model training and inference,3D character and scene automatic generation,inference computation acceleration and lowresource hardware adaptation,and generative embodied environment perception with interaction techniques.Empirical evidence indicates that this technology has demonstrated significant effectiveness in military scenarios such as dynamic perception of battlefield situations,multi-source intelligence fusion and reorganization,and intelligent decision support.Building on the above findings,it is urgent to promote the integration of national strategic planning,technological capability supply,basic support system,and construct a novel paradigm of technological development oriented towards intelligent application needs,in order to achieve strategic breakthroughs and sustainable innovation of Generative AI in the military field.
作者
王亚珅
郝浚霄
刘振宇
孙新
陈天柱
WANG Yashen;HAO Junxiao;LIU Zhenyu;SUN Xin;CHEN Tianzhu(Cyber-Electromagnetic Space Intelligent Laboratory(CESIL),Artificial Intelligence Institute of CETC,Beijing 100049,China;School of Computer,Beijing Institute of Technology,Beijing 100081,China)
出处
《无人系统技术》
2026年第1期137-147,共11页
Unmanned Systems Technology
基金
国家自然科学基金(U22B2061)。
关键词
生成式人工智能
人工智能
多模态大模型
3D生成
轻量化
具身智能
Generative Artificial Intelligence
Artificial Intelligence
Multi-modal Large Model
3D Generation
Lightweight
Embodied Intelligence
