通信世界网消息（CWW）随着人工智能技术的加速发展，生成式AI驱动的创新应用场景不断涌现，对 算力和网络基础设施提出了新的挑战。智能算力是支撑生成式AI的核心资源，而高效的算力互联则可 以发挥集群效应，进一步拓展智算性能上限和服务边界。同时，智能体作为基于生成式AI的新服务形 态，其泛在的协作通信成为IP广域网发展的重要方向之一。这些因素促进IP广域网从"连接赋能"到"算 力赋能"升级，并正在迈向"AI赋能"新阶段。 国家相继发布政策支持算力及AI技术发展。2023年，我国发布《算力基础设施高质量发展行动计划》 《关于深入实施"东数西算"工程加快构建全国一体化算力网的实施意见》等，强调算力高质量发展、算 网融合基础设施建设的重要性。2025年，国务院《关于深入实施"人工智能+"行动的意见》进一步强调 了强化智能算力统筹，完善全国一体化算力网，充分发挥"东数西算"国家枢纽作用，加强数、算、电、 网等资源协同。 在国家政策和产业需求双轮驱动下，需要构建AI内生的高智能、高可靠、高安全的IP广域网络，智能IP 广域网（AI WAN）应运而生。AI WAN基于算力网络新型基础设施与AI深度融合，通过构建新的AI网 络 ...

正值大模型从"算力堆叠"走向"推理优先"的关键节点，清华大学智能产业研究院（AIR）创始院长、中国工程院外籍院士张亚勤提出： 新一轮人工智能，是信息智能、物理智能和生物智能的融合，本质上也是原子、分子和比特的融合。 也就是说，在规模定律持续发挥作用的前提下，当参数规模、数据体量与算力资源跨过某个阈值，智能就不再只停留在模式识别，而是开始"涌现"—— 先是从鉴别式 AI 走向生成式 AI，再从生成式 AI 走向以智能体为代表的新范式。 在本次量子位 MEET2026 智能未来大会上，他也将 ChatGPT 和 DeepSeek，视作这一轮演进中的两个重要里程碑： 从 ChatGPT 到 DeepSeek，AI 正沿着"智能 +"的路径进入新一轮浪潮。 至于未来 5~10 年的主战场，在他看来，将走向"智能体互联网"时代 —— 基础大模型像操作系统一样在全球范围内收敛到不超过 10 个；而智能体会取代 今天的大部分 SaaS 和 App，成为企业和个人与世界交互的默认形态，同时这也是通往 AGI 的必经之路。 为了完整体现张亚勤的思考，在不改变原意的基础上，量子位对演讲内容进行了编辑整理，希望能给你带来更多启发。 ...

正值大模型从"算力堆叠"走向"推理优先"的关键节点， 清华大学智能产业研究院（AIR）创始院长、中国工程院外籍院士张亚勤 提出： 编辑部 整理自 MEET2026 量子位 | 公众号 QbitAI 从ChatGPT到DeepSeek，AI正沿着 "智能+" 的路径进入新一轮浪潮。 新一轮人工智能，是信息智能、物理智能和生物智能的融合，本质上也是原子、分子和比特的融合。 也就是说，在规模定律持续发挥作用的前提下，当参数规模、数据体量与算力资源跨过某个阈值，智能就不再只停留在模式识别，而是开 始"涌现"—— 先是从鉴别式AI走向生成式AI，再从生成式AI走向以 智能体 为代表的新范式。 在本次 量子位MEET2026智能未来大会 上，他也将ChatGPT和DeepSeek，视作这一轮演进中的两个重要里程碑： 前者通过统一表征与token化，把文本、语音、图像乃至蛋白质、点云等数据纳入同一空间； 后者则以高效率、高性能、低价格和开源路径，把大模型从"预训练时代"推向以推理为核心的"DeepSeek时刻"。 至于未来5~10年的主战场，在他看来，将走向 "智能体互联网"时代 —— 基础大模型像操作系统一样在全球范围内 ...

前几天，一位朋友问我："OpenAI的AI助手已经能直接控制智能家居了，谷歌的Gemini也在接管汽车，我们这些做物联网硬件的是不是要被短路了？" 我说：恰恰相反，AI越强大，物联网越重要。 最近"智能体互联网"这个概念最近频繁出现在各种讨论中，但大多数人都在关注上层的AI大模型，却忽略了一个关键事实：没有物联网，智能体就是缸中 之脑；没有智能体，物联网就是植物人。 我们正在见证的，不是物联网被AI短路，而是物联网从前台明星退居幕后基建的关键时刻。就像互联网时代没人再讨论TCP/IP协议，但它无处不在，物 联网正在经历同样的下沉，成为智能体时代的具身数字底座。 智能体互联网，本质上是让万物从被动的数据终端进化为能主动思考、相互协作并直接交付结果的智能伙伴。这不是简单的技术迭代，而是硅基智能 从"观察世界"向"改变世界"的惊险一跃。 从连接到智能：一场三十年的数字进化史 回头看这三十年，我们经历了五次关键的演进，每一次都在解决前一次留下的缺憾。更有意思的是，这些看似独立的技术浪潮，其实都在朝着同一个方向 努力，让数字世界和物理世界的边界彻底消失。 第一幕发生在1990年代，互联网完成了信息的数字化。 那个年代， ...

现场演示引爆关注：AI如何真正"赋能"而非"替代" 演讲现场，老蒋通过实时演示，展示了AI智能体在IP定位诊断、爆款脚本生成、朋友圈内容策划、心理 测评、皮肤检测等多个场景下的应用效果。其中，"AI定位大师"在5分钟内完成一场品牌定位沟通，并 生成结构化报告；"爆款文案智能体"可根据用户需求实时生成情感饱满、符合人设的口播脚本，引发全 场赞叹。 老蒋透露，创客匠人内部已全面推行"组织AI化"，全员具备编写AI智能体的能力，并初步实现了"人均 管理多个数字员工"的运营模式。公司计划在两年内，推动每位员工管理上百个智能体，从而实现"400 名员工、4万名数字员工"协同作业的生态愿景。 2025年11月22日至25日，由创客匠人主办的"全球创始人IP+AI万人高峰论坛"在厦门海峡大剧院圆满落 幕。本届大会以"AI、IP、未来、共创、格局"为关键词，汇聚了新商业架构师张琦、梅花创投创始合伙 人吴世春、北京师范大学教授于丹、能力风暴创始人恽为民、前字节跳动市场负责人叶旭东、AI科技 公司创始人严伯钧等众多商业领袖与行业专家，吸引了全球超万名创始人IP共聚一堂，共同探讨AI时代 下个人品牌与商业增长的新路径。 作为大会主 ...

Core Insights - The report discusses the development of artificial intelligence (AI) computing centers and their evolution into AI computing networks, emphasizing the importance of infrastructure for digital economy growth and technological advancement [6][7][10]. Group 1: Global AI Computing Center Development - AI computing centers are recognized as critical infrastructure for the digital economy, facilitating the integration of AI technologies with traditional industries [10][18]. - Major countries, including the US, Europe, Japan, and China, are actively investing in AI computing infrastructure to maintain global competitiveness [12][18]. - The US plans to invest approximately $69 billion in AI and quantum computing research and infrastructure over the next five years, aiming to triple the number of AI research institutions by 2025 [13][19]. Group 2: China's AI Computing Center Characteristics - China's AI computing centers are primarily government-led, focusing on autonomous technology development and market-oriented operational mechanisms [31][33]. - The construction of AI computing centers is seen as a vital support for the integration of AI with the real economy, with cities like Wuhan and Shenzhen leading the way in establishing operational centers [23][34]. - The report highlights the importance of creating a public service platform that aggregates innovation resources and empowers industrial development through AI computing centers [46]. Group 3: New Demands and Challenges - The demand for computing power in AI is increasing, driven by large-scale pre-trained models that are being applied across various industries [7][23]. - AI computing centers must adapt to the new landscape characterized by "large models + big data + high computing power" and align with national carbon reduction goals [7][31]. - The report suggests that AI computing centers should evolve into interconnected computing networks to enhance resource sharing and meet the new demands of the economy [7][8]. Group 4: Recommendations for High-Quality Development - The report recommends a coordinated approach to enhance the quality of AI computing center construction and to establish operational mechanisms for AI computing networks [7][8]. - It emphasizes the need for technological innovation and policy support to facilitate collaboration between academia, industry, and government [7][8]. - Standardization in the construction of AI computing networks is crucial for their sustainable development and scalability [7][8].

Core Insights - The internet industry in China generated 14,420 billion yuan in revenue in the first three quarters of this year, marking a 2.8% year-on-year growth, indicating its role as a driving force for economic development [1] - Artificial intelligence (AI) is identified as a significant engine for the transformation of the internet industry, with the core AI industry exceeding 900 billion yuan, leading global open-source innovation ecosystems [1] - The emergence of intelligent agents signifies a new development phase in the internet, where AI evolves from a mere tool to an independent digital entity capable of perception, decision-making, and collaboration [2] Industry Developments - The intelligent agent internet expands the network's connectivity from humans and devices to intelligent agents, establishing a new norm of human-machine coexistence and intelligent collaboration [2] - Companies are actively developing intelligent agent products, enhancing work efficiency and enabling multi-step process automation across business departments [2] - New applications like "AI Smart Search" and "AI Market Trends" were launched by Xianyu, utilizing AI to improve transaction efficiency and provide users with tailored product recommendations based on real-time data [3] AI Application and Integration - AI technology is transitioning from concept to large-scale application, significantly altering the operational dynamics of various sectors, including manufacturing and services [4] - The integration of large models is pushing industries from "productivity enhancement" to "intelligent reconstruction," reshaping interaction and knowledge logic [4] - Companies are focusing on embedding industrial knowledge into model training to ensure flexible technology deployment across various sectors [5] Collaborative Efforts and Challenges - The deep application of large models and intelligent agents raises security risks that require collective attention from all stakeholders [6] - A dual governance strategy combining external and internal security measures is proposed to address the complexities of AI security [6] - Legal and policy frameworks are essential for balancing development and security in the internet sector, ensuring that advancements benefit society [6][7]

Core Insights - The 2025 World Internet of Things Expo marks the beginning of a significant transformation in the industry, focusing on the evolution from "connecting everything" to "intelligent interconnection" and ultimately to "intelligent action" [3][20] - The event emphasizes the importance of collaboration and capability development, showcasing a comprehensive AIoT capability map that outlines the future of the industry [3][20] Domain Expansion - The expo provides a clear answer to the question of how to achieve endless expansion of the AIoT industry, emphasizing the need for authoritative results to measure and define the industry [4][5] - The release of the "IoT Industry Scale Measurement Index System" offers a quantifiable way for the industry to assess its value, marking a significant step in defining the economic landscape of "intelligent action" [5] - The event has achieved unprecedented international participation, with nearly 400 companies from 15 countries, indicating a growing global influence of the AIoT capabilities represented by the expo [6] Application Scenarios - The expo showcases a variety of application scenarios, illustrating the journey from "intelligent perception" to "network connection" and "vertical empowerment" [9][12] - Core capabilities are highlighted as the foundation for application explosion, with advancements in multi-modal sensing and communication technologies being showcased [11] - The event emphasizes the importance of diverse ecosystems, exploring new application domains beyond traditional industries, such as emergency logistics and environmental monitoring [13] Future Aspirations - The expo serves as a platform for global consensus and vision, with high-level meetings and expert consultations shaping the future of the AIoT industry [16][17] - Innovation is positioned as a key driver for the industry's future, with a focus on core technologies and support for developers and entrepreneurs [17][18] - Collaborative ecosystems are being built to connect various stakeholders, facilitating resource matching and accelerating the development of the regional industrial chain [18][19]

Core Viewpoint - The Internet of Agents (IoA) represents a new digital ecosystem composed of autonomous agents that can discover, communicate, and collaborate without direct human intervention, marking a shift towards an "agent-driven" internet model [1][2]. Group 1: Market Potential and Growth - The IoA is rapidly gaining traction due to technological breakthroughs and increasing demand for automation, with Gartner predicting that by 2028, 15% of daily tasks will be autonomously handled by AI agents, and about one-third of enterprise applications will embed agent intelligence [2]. - The market for agent-related technologies is expected to grow at a remarkable compound annual growth rate (CAGR) of 46% in the coming years, indicating its potential to disrupt business models and lifestyles [2]. - By 2035, it is estimated that the global number of AI agents will reach 900 billion, with over 90% of Chinese households expected to own smart robots, highlighting the scale of IoA's future ecosystem [2]. Group 2: Technical Foundations and Challenges - The evolution of the IPv6 protocol is crucial for the IoA, addressing challenges such as addressing, dynamic networking, and security verification for large-scale agent collaboration [3][4]. - The IoA requires a vast address space to uniquely identify billions of agents, necessitating a new naming service discovery system that can dynamically register and parse semantic rules [3][4]. - Security and trust are fundamental for autonomous collaboration among agents, with IPv6 enabling a comprehensive security framework for identity, connection, and data [4]. Group 3: Data Governance and Utilization - Data is the foundation of the IoA, with agents processing vast amounts of heterogeneous data, necessitating robust data sovereignty and compliance measures to prevent misuse [5][6]. - A distributed data collaboration platform is essential for ensuring secure, compliant, and controllable data usage and exchange among agents, facilitating efficient service discovery and task completion [6][7]. - The data space establishes clear rules for data usage, ensuring accountability and traceability, which is vital for the operational integrity of the IoA [6][7]. Group 4: Strategic Development and Recommendations - China has a unique opportunity to develop the IoA, supported by a robust IPv6 infrastructure, a large data resource pool, and strong government backing for digital transformation [8][9]. - Recommendations for advancing the IoA in China include formulating a national development strategy, enhancing legal frameworks, building an innovation ecosystem, and fostering international cooperation [9][10]. - The development of the IoA is seen as both a challenge and an opportunity for industrial upgrading, with the potential for significant breakthroughs if innovation and market advantages are leveraged effectively [10].

Core Viewpoint - The article discusses the evolution of policy from "Internet Plus" to "Artificial Intelligence Plus," highlighting the differences in focus and implications for the future of technology integration in China [2][5][18]. Policy Evolution - The "Artificial Intelligence Plus" action plan was released by the State Council on August 26, 2025, aiming to deeply integrate AI with various sectors of the economy and society, outlining six key actions and corresponding developmental goals [2][3]. - In contrast, the "Internet Plus" action plan was issued on July 4, 2015, emphasizing the role of the internet as a foundational infrastructure for resource integration and innovation [3][7]. Terminology and Focus - A notable difference between the two documents is the frequency of the term "platform," which appears 62 times in the "Internet Plus" document but only 4 times in the "Artificial Intelligence Plus" document, indicating a shift in focus from platform-centric models to a broader integration of AI [4][5]. Technological and Economic Differences - The article argues that the fundamental nature of the internet as a "connector" contrasts with AI as a "cognitive" technology, leading to different policy approaches. The internet's value lies in connecting dispersed entities, while AI enhances the intelligence of individual nodes [9][10]. Industry Practices - The concept of "industrial internet" emerged from the "Internet Plus" initiative, where platform companies sought to extend their influence from consumer to enterprise sectors. However, the reality of business needs has led to a decline in enthusiasm for this strategy [11][12]. AI Integration and Future Prospects - The "Artificial Intelligence Plus" initiative aims to transition service industries from digital empowerment to intelligent-driven services, suggesting a shift in how businesses will adopt AI technologies [13][14]. - The article expresses optimism about achieving the 2027 goal of widespread AI integration across six key areas, citing lower resistance compared to the "Internet Plus" initiative [13][14]. Long-term Goals and Network Strategies - The long-term objectives for 2030 and 2035 require a network strategy to support the comprehensive development of the intelligent economy, emphasizing the need for interaction between supply and demand [16][18]. - The article notes a lack of emphasis on the internet within the "Artificial Intelligence Plus" policy, which may overlook the potential benefits of network effects in facilitating AI's growth [17][18]. Intelligent Internet Concept - The concept of "intelligent internet" is introduced as a potential framework for integrating AI and internet technologies, suggesting that the evolution from platform models to intelligent agents could enhance the effectiveness of AI applications [19][20]. - The emergence of open-source protocols for intelligent agents may facilitate a transition to a more decentralized and efficient model, breaking the dominance of traditional platform companies [19][20]. Conclusion - The article concludes that both "Internet Plus" and "Artificial Intelligence Plus" are not mutually exclusive but rather interdependent, with the potential to jointly drive China towards a new stage of intelligent economic and social development [22].