大模型时代:高质量数据的构建、治理与未来展望  

The Age of Big Models:Building,Governing and Exploring of High-Quality Data

作　　者：穆晓君 吴桐 加雄伟[1] 马瑞涛[1] Mu Xiaojun;Wu Tong;Jia Xiongwei;Ma Ruitao(China Unicom Research Institute,Beijing 100048,China;China National Building Materials Group Corporation Finance Company Limited,Beijing 100036,China)

机构地区：[1]中国联通研究院,北京100048 [2]中国建材集团财务有限公司,北京100036

出　　处：《信息通信技术》2024年第3期63-68,共6页Information and communications Technologies

摘　　要：在当前人工智能技术飞速发展的背景下,大模型以其强大的数据处理能力和复杂任务执行能力,逐渐成为推动各行业创新与变革的核心动力。数据质量被认为是决定大模型性能表现的关键要素之一,其重要性日益凸显。文章旨在探讨高质量数据如何驱动大模型能力,分析高质量大模型数据的特征,提出构建高质量数据的关键流程,探讨新时代大模型数据治理方向,并对高质量数据未来趋势进行展望,以期为政策制定者、行业参与者及研究人员提供洞见参考。With the rapid development of artificial intelligence technology,big models have gradually become the core driving force for innovation and change in various industries due to their powerful data processing capabilities and complex task execution capabilities.Among them,data quality is regarded as the key factor to determine the performance of large models,and its importance is becoming increasingly prominent.This paper aims to explore how high-quality data drives large model capabilities,analyze the characteristics of high-quality large model data,propose key processes for building high-quality data,explore the direction of large model data governance in the new era,and look forward to the future trend of high-quality data,in order to provide insights for policy makers,industry participants and researchers.

关 键 词：数据质量 大模型 数据治理 人工智能 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP18[自动化与计算机技术—计算机科学与技术]