摘要
在生成式人工智能大模型爆炸式发展背景下,智能计算的重要性日益凸显。当前千亿、万亿级参数大模型在海量数据训练过程中,对智能计算在芯片算力、内存容量、互联速率等方面提出更高要求。从支持大模型创新发展需求入手,重点从芯片技术、软件技术、互联技术等维度分析智能计算技术发展态势和产业发展现状,结合当前产业发展面临的机遇与挑战,提出未来智能计算发展策略。
Against the backdrop of the explosive development of generative Artificial Intelligence(AI)large language models,the importance of intelligent computing is becoming increasingly prominent.Currently,large language models with hundreds of billions or even trillions of parameters have raised demanding requirements for intelligent computing in terms of chip computing power,memory capacity,and interconnection speed during the training process of massive data.Starting from the demand for key technologies of intelligent computing,this study focuses on analyzing the development trend and the current industry situation,in dimensions such as computing chip technology,software technology,and interconnection technology.Combining with the opportunities and challenges faced by the current industry development,it proposes future development strategies for intelligent computing.
作者
张乾
黄璜
苗子聪
ZHANG Qian;HUANG Huang;MIAO Zicong(Informatization and Industrialization Integration Research Institute,China Academy of Information and Communications Technology,Beijing 100191,China;Cloud&Network Product Division,China Telecom Cloud Computing Corporation,Beijing 100083,China)
出处
《信息通信技术与政策》
2024年第6期10-16,共7页
Information and Communications Technology and Policy
