生成式人工智能赋能国防科技情报  

Generative AI Empowers National Defense S&T Information Work

作　　者：汤珊红 李晓松 赵柯然 耿国桐 Tang Shanhong

机构地区：[1]军事科学院军事科学信息研究中心,北京100142

出　　处：《情报理论与实践》2023年第11期81-85,99,共6页Information Studies:Theory & Application

摘　　要：[目的/意义]发掘生成式人工智能技术在国防科技情报领域应用的潜在价值,探索人工智能赋能国防科技情报工作转型的方法对策。[方法/过程]从国防科技情报领域的核心关切出发,结合主流大语言模型主体架构,从语料体系、预训练算法与模型、微调算法与模型三个层次归纳了对生成式人工智能的认识;立足国防科技情报工作的本质特征,从情报收集、情报评估、情报分析、情报生成环节分析了生成式人工智能在国防科技情报工作领域的潜在应用;针对生成式人工智能给国防科技情报工作带来的反情报工作风险、循证能力问题、准确性、时效性、安全性等挑战,提出了有效利用生成式人工智能的对策建议。[结果/结论]国防科技情报领域从业者应当积极融入新技术大潮,从受益者转变为贡献者。[Purpose/significance]This study explores the potential value of generative AI technology in the field of national defense S&T information,and explores methods and strategies for empowering the transformation of national defense S&T information work with AI.[Method/process]Starting from the core concerns in the field of defense S&T information,this study combines the main architecture of mainstream large language models to summarize the understanding of generative AI from three levels:corpus system,pre-training algorithms and models,and fine-tuning algorithms and models;Based on the essential characteristics of national defense S&T information work,this paper analyzes the potential applications of generative AI in the field of national defense S&T information work from the aspects of information collection,evaluation,analysis,and generation;And in response to the challenges posed by generative AI in counterintelligence work,evidence-based capabilities,accuracy,timeliness,and security,effective utilization of generative AI is proposed.[Result/conclusion]Practitioners in the field of defense S&T information work should actively integrate into the new technology trend,transforming from beneficiaries to contributors.

关 键 词：生成式人工智能 大语言模型 国防科技情报 情报循证 情报生产线 

分 类 号：G350[文化科学—情报学]