2026.03. 03 本文字数：734，阅读时长大约2分钟 封图 | 受访者供图 2026年全国两会即将召开，全国政协委员、360集团创始人周鸿祎拟围绕推理算力、智能体技术与 人才发展、智能安全三方面提交提案。 周鸿祎认为，我国经历"百模大战"后，有了很多"国际一流"的开源模型，国家主导的训练算力稳步提 升，推理算力需求在"百亿智能体时代"呈指数级增长，专用推理芯片是我国芯片产业实现差异化突围 的重要方向。 目前，我国算力中心面向推理任务的专用集群存在缺口，区域间供需适配有待优化，专用推理芯片技 术也亟需突破。周鸿祎建议优化推理算力布局：国家出台推理算力布局指导政策，依据各地场景密 度、算力缺口、能源保障能力，建立"全国统筹 + 区域细化" 的推理算力布局体系，在重点产业集聚 区域，建设低时延、高密度的推理算力集群。 另外，强化一体化调度，推动跨层级、跨区域的算力资源动态调配，提升推理算力利用效率。鼓励专 用推理芯片的国产化发展，重点突破高精度、低时延、多模态的芯片技术，实现产业链自主可控，支 持智能体技术的深度应用。 采访中，周鸿祎表示，国产大模型未达到及格线时，重视训练算力是合理的，但如今极度消耗推理算 ...

Core Viewpoint - The integration of AI technology with office scenarios presents a significant opportunity for the industry, aligning with the company's strategic focus on intelligent agent technology, which has been in development for three years [1] Group 1: Company Strategy and Development - The company has established the Rich AIBox integrated platform, which is entering a critical phase of large-scale implementation, aligning with industry trends of "agent collaboration, safety and control, and human-machine collaboration" as reported by Anthropic [1] - The company's methodology outlined in the "Enterprise Service AI White Paper" emphasizes that the value of enterprise AI equals safety compliance multiplied by (scenarios + data + models + services), showcasing its forward-looking approach [1] - The company is committed to "continuous investment and pragmatic implementation" in the intelligent office sector, with a projected 20.13% year-on-year increase in R&D investment for 2024, focusing on optimizing the "Rich AI full-stack service system" [1] Group 2: Product Development and Market Application - The company has developed core office products like RichMail, which have been adapted for HarmonyOS, providing technical support for products with hundreds of millions of users, such as those from China Mobile [1] - Future plans include expanding into multi-terminal scenarios and customized industry solutions, following the white paper's guidance to target high-frequency essential scenarios, integrating intelligent agent technology into collaborative office and knowledge management core areas [1] - The company will maintain compliance in operations, focusing on technological R&D and product innovation, with the Rich AIBox platform at its core, continuously transforming intelligent agent capabilities into perceivable business value for clients [1]

AI（人工智能）会终结软件行业？围绕这一热门话题，近日，市场和行业纷纷用行动投票。美国华尔 街对AI冲击软件业的担忧持续发酵，软件股遭遇抛售，抛售潮迅速蔓延至全球市场。与此同时，科技 界发出截然不同的声音。英伟达CEO黄仁勋直言，"AI将取代软件工具"的观点"极不合逻辑"；OpenAI 首席执行官奥尔特曼也表示，软件的创造方式、使用方式和商业模式正在发生深刻变化，但"软件不会 消失"。 引发这波争议的导火索，是AI初创企业Anthropic推出的垂直工具对传统软件功能的冲击：这款AI法务 插件能够执行多项文书工作，包括追踪合规事项、审阅法律文档，而这些正是许多法律软件产品的核心 功能。当AI能够自动完成大量原本依赖专业软件的工作时，市场不免疑问：如果AI可以直接完成任 务，我们还需要单独的软件吗？ 这种悲观情绪从单一AI工具的冲击，逐渐演变为对软件行业整体商业模式的重估，并最终指向一个核 心问题——SaaS（软件即服务）的商业模式是否正在被AI动摇？ 过去十多年，订阅收费、持续更新、客户黏性强的SaaS模式，一直是软件产业最稳定的增长引擎。但如 今，这一切不再"顺理成章"。随着生成式AI和智能体技术快速发展， ...

Group 1 - The core event was the "Cloud Intelligence Integration · Intelligent First" Agent Conference held by Baidu Smart Cloud, where Baidu Qianfan announced its achievements, including support for over 1.3 million agents and daily tool usage reaching tens of millions [1] - Baidu Qianfan has integrated agents into enterprise production lines, moving beyond experimental applications to support innovation in key industries such as smart hardware, manufacturing, transportation, and energy [1] - The launch of the "Baidu Qianfan Developer Ambassador Program" aims to collaborate with experienced developers to conduct training camps and activities nationwide, accelerating the large-scale implementation of enterprise-level intelligent agents [1] Group 2 - Li Jingqiu, Chief Architect of Baidu Smart Cloud AI and Large Model Platform, stated that by 2025, the industry will complete the paradigm shift from validation to implementation, with a significant explosion of "native agents" expected in 2026 [2] - Baidu Qianfan is addressing cost and threshold challenges faced by enterprises by building a comprehensive enterprise-level Agent Infrastructure that includes model services, tool services, agent development services, data services, and operational environments [2] - A new enterprise-level long memory capability MVP version was released, enabling cross-agent memory sharing and intelligent extraction, allowing agents to have "thinking continuity" over time and become digital assets that understand business and user preferences [2]

Group 1 - The company expects a net profit for the year 2025, but it will decline by over 50% compared to the previous year [1] - The performance forecast period is from January 1, 2025, to December 31, 2025 [1] - The company has communicated with its auditing firm regarding the performance forecast, and there are no significant discrepancies [1][3] Group 2 - The company's overall operating revenue remained stable compared to the previous year, with growth in the smart rail business due to planned order deliveries [3] - The company is advancing its "AI+" development strategy, increasing investment in AI, large models, and intelligent technology for education and rail scenarios, which has led to an increase in sales, R&D, and management expenses by approximately 33 million yuan [3] - Improved management of accounts receivable has resulted in a significant reduction in credit impairment losses for two consecutive years, positively impacting operational quality and net profit [3] Group 3 - The company plans to continue focusing on the deep integration of AI technology with industry scenarios and accelerate the large-scale implementation of innovative results [4]

Group 1 - The core point of the article is that Sairisi Group Co., Ltd. has applied to register three trademarks for "Yixin Intelligent Body," which are currently in the status of awaiting substantive examination [1] - Sairisi Group was established in May 2007 and has a registered capital of approximately 1.63 billion RMB [1] - The company's business scope includes manufacturing and selling automotive parts, motor vehicle parts, general machinery, electrical machinery, and electronic products [1] Group 2 - The shareholders of Sairisi Group include Chongqing Xiaokang Holdings Co., Ltd., Dongfeng Motor Group Co., Ltd., and Chongqing Yuhan Automotive Industry Co., Ltd. [1]

Core Insights - Zhongke Shuguang (603019) announced that the National Industrial Information Security Development Research Center published the "2024-2025 China Intelligent Agent Application Research Report" [1] - The report includes the intelligent agent services provided by Shuguang Cloud for the Gansu Province "5G + Smart Highway" cloud platform project, which supports "intelligent decision-making in highway traffic construction" [1] - This service represents a significant application of intelligent agent technology in the field of transportation infrastructure [1]

Group 1 - South Korean President Lee Jae-myung will make his first state visit to China from January 4 to 7, 2026, marking him as the first foreign head of state to visit China in the new year [1] Group 2 - The recruitment process for mainland universities to accept graduate students from Hong Kong and Macau has started on January 4, 2026, with candidates required to register on the "Xuexin Network" and submit their applications online [2] - Each candidate is limited to applying for one program at one university, and late applications or modifications to submitted information are not permitted [2] Group 3 - The 2026 National Intelligent Agent Developers Conference will be held in Changzhou on January 4, 2026, aimed at promoting the integration of artificial intelligence technology into local industries [3]

Core Viewpoint - The 2026 National Intelligent Agent Developers Conference is a significant initiative to promote the integration of artificial intelligence technology into local industries and is a crucial step for Changzhou to capitalize on the emerging intelligent agent technology trend [1] Group 1 - The conference is set to take place on January 4, 2026, in Changzhou [1] - It aims to empower local industries through the adoption of artificial intelligence technologies [1] - The event reflects Changzhou's strong industrial foundation and its strategic move to seize opportunities in the intelligent agent technology sector [1]

Core Insights - The launch of the Doubao mobile assistant in late 2025 has created significant waves in the smartphone industry, being recognized by users as a "truly AI phone" due to its autonomous cross-application operation capabilities [1][5] - However, the Doubao mobile assistant faced usage restrictions in multiple applications, leading to an official statement on December 5 regarding adjustments to its AI operation capabilities [1][3] - The challenges faced by Doubao reflect broader issues in the AI phone sector, particularly concerning user privacy and data rights, which require collaborative efforts from various stakeholders [3][7] Group 1: Technology and Development - The rapid improvement of domestic open-source large models has provided a solid foundation for the iterative development of AI phones, shifting focus from cloud-based models to offline capabilities [4][13] - The announcement of the open-sourcing of the AutoGLM model by Zhiyu on December 9 signifies a move towards "technological equality" in the mobile agent space, potentially lowering development barriers [3][12] - The evolution of AI phones is not just about technological advancements but also involves a complex interplay of patience, restraint, and ecosystem collaboration [4][17] Group 2: Market Dynamics and Competition - The Doubao mobile assistant's features are not significantly different from previously demonstrated smart applications by other manufacturers, indicating that the market is still in the early stages of realizing cross-application autonomous capabilities [5][6] - The competition in the AI phone market is fundamentally about ecosystem development, requiring a shift from traditional app models to a more integrated approach that allows for cross-application automation [17][21] - The emergence of AI agents is seen as a potential battleground for user engagement and control over data, with manufacturers needing to navigate the complexities of user habits and regulatory compliance [8][17] Group 3: Industry Challenges - The current landscape reveals a "standard vacuum" regarding user authorization and data rights, complicating the implementation of AI agents that require cross-application functionality [7][18] - The transition to AI agents necessitates a reevaluation of existing operational models, as traditional single-operation authorization methods are inadequate for the demands of AI-driven tasks [7][18] - The industry faces challenges related to memory and power consumption as the capabilities of end-side models expand, necessitating advancements in hardware to support these new demands [18][20] Group 4: Future Outlook - The AI phone competition is evolving from a focus on hardware specifications to a comprehensive race involving end-side capabilities, ecosystem openness, and user experience [21][22] - The integration of AI capabilities into hardware is expected to reshape traditional consumer electronics business models, moving towards subscription-based software services [20][22] - The developments surrounding the Doubao mobile assistant highlight the need for industry-wide collaboration to address the challenges of rights, interests, and standards in the evolving AI landscape [20][22]