专访信通院孙鑫：大模型快速迭代需软硬件深度协同

Core Insights - The Chinese government emphasizes the importance of standards in promoting high-quality economic development, particularly in the context of artificial intelligence and digital technologies [1] - The Ministry of Industry and Information Technology highlights China's commitment to high-level opening-up and the advancement of "smart industrialization" and "industrial intelligence" [1] - The development of artificial intelligence is marked by several key trends, including the deep collaboration between hardware and software, the emergence of intelligent agents, and the acceleration of model iteration [2][3] Group 1: Trends in Artificial Intelligence - The integration of hardware and software is becoming a new paradigm for developing large models, with extreme collaboration being crucial for rapid iteration [3] - Intelligent agents are emerging as the primary form of large model applications, contributing to the formation of an intelligent economy [3] - The rapid iteration of foundational large models is evident, with a 90% overall improvement in multimodal model understanding capabilities since last year [2][3] Group 2: Intelligent Agents and Their Development - Intelligent agents, as the initial form of digital employees, are capable of autonomously completing complex tasks, although there is still significant room for improvement [4][10] - The development of intelligent agents is characterized by the need for enhanced interconnectivity and the ability to handle long-duration tasks [10][11] - Communication protocols are essential for expanding the capabilities of intelligent agents and addressing data silos [10] Group 3: Industry Applications and Challenges - The penetration of artificial intelligence across different industries varies, with a tendency for initial breakthroughs in sectors with higher digitalization levels [12][13] - Industries such as finance, healthcare, and transportation are seeing significant advancements in AI applications, particularly in autonomous driving [13] - The need for coordination between industry levels and transformation routes, as well as between technical capabilities and actual demands, is critical for successful AI implementation [12][13]