展期论坛 | AIE智算生态大会，议程及嘉宾阵容抢先看

算力已成为智能时代发展的核心驱动力，如何构建高效协同的算力网是算力基础设施化的关键技术。本报告将结合中国算力网（C2Net）研究计划的实践经验，介绍对算力网软硬件技术发展现状、核心挑战与未来趋势的分析与思考。

Computility has become the core driving force for development in the intelligent era, and building an efficiently collaborative computing network constitutes a key technology for the infrastructureization of computing resources. Drawing on practical experience from the China Computing Net (C2Net) research initiative, this report presents analyses and reflections on the state-of-the-art, core challenges and future trends of hardware and software technologies for the computing network.

本次演讲围绕“交通时空原生大模型探索和智能体应用实践”这一主题，介绍深城交在交通行业大模型方向的探索。演讲首先从交通行业近年来数据基础、感知能力、计算能力和业务数字化水平持续提升的背景出发，结合实际业务中逐步出现的需求，引出大模型在交通领域进一步发挥作用的空间。在此基础上，介绍深城交如何依托既有交通数据、专业算法模型、行业知识和业务系统，探索构建面向交通时空任务的大模型能力，并进一步通过智能体实现任务理解、能力调用、分析推演和业务协同。演讲将结合综合交通规划、常态运行、重大活动保障、突发事件处置等典型场景，展示相关技术在实际业务中的应用方式与阶段性成效。最后，进一步探讨大模型和智能体如何与算力、数据、模型工具、业务系统及交通设施协同演进，推动交通AI能力由项目化应用走向可复用、可组合、可持续演进的行业智能基础设施。

This presentation focuses on the theme of "Exploration of Transportation Spatiotemporal Native Large Models and Agent Application Practices," introducing SUTPC's exploration in the direction of large models for the transportation industry. The presentation first starts from the background of continuous improvements in recent years in data foundations, perception capabilities, computing power, and business digitalization levels in the transportation industry, and, in combination with needs gradually emerging in practical business operations, identifies the space for large models to play a further role in the transportation field. On this basis, it introduces how SUTPC, relying on existing transportation data, specialized algorithm models, industry knowledge, and business systems, explores the construction of large model capabilities oriented toward transportation spatiotemporal tasks, and further enables task understanding, capability invocation, analysis and deduction, and business collaboration through agents. Combining typical scenarios such as comprehensive transportation planning, routine operations, major event support, and emergency incident response, the presentation will demonstrate the application methods and phased results of relevant technologies in actual business operations. Finally, it further discusses how large models and agents can co-evolve with computing power, data, model tools, business systems, and transportation facilities, promoting the evolution of transportation AI capabilities from project-based applications toward reusable, composable, and sustainably evolving industry intelligent infrastructure.

AI发展已进入以“Token”为核心衡量单位的经济时代，中兴通讯致力于通过全栈协同，深耕多芯协同、架构创新、连接增强、持续演进等核心能力，打造TCO最优的AI基础设施，让成本更优、效率更高、体验更好，从而消除Token焦虑，助推AI普惠。

ZTE is committed to full-stack open co-design to build AI infrastructure with optimal TCO, delivering superior cost, efficiency and experience. It alleviates Token pressure and advances inclusive AI.