（文章来源：南方都市报） 在刚刚发布的三季度财报里，百度首次披露AI业务收入，达到约100亿元，同比增长超50%。其中，AI 云收入同比增长33%，达到42亿元；AI应用收入达26亿元；AI原生营销服务收入同比增长262%达到28 亿元。 根据财报，本季度长期资产减值为162亿元，主要因核心资产集团的减值损失。对于这次的资产减值， 百度高管在财报电话会上透露，百度正在毫不犹豫地加大对最新AI计算技术的投资，"我们对基础设施 组合进行了全面审查。一些现有资产已不再满足计算效率要求，因此，我们主动进行了一些减值处理。 经过这次一次性减值后，我们的资产基础和组合状况更加健康，更好地与先进的AI计算需求和高价值 应用场景相契合。" 在资本支出方面，百度高管透露，自2023年3月发布文心一言以来，公司在AI投资上的投资已经远远超 过了1000亿元人民币。展望未来，公司将继续加大在AI领域的投资力度。 11月13日，百度正式发布原生全模态大模型文心大模型5.0。该模型采用原生全模态统一建模技术并应 用超稀疏混合专家架构，支持理解与生成一体化，具备全模态理解与生成能力，支持文本、图像、音 频、视频等多种信息的输入与输出。 ...

11月18日，百度发布第三季度财报，首次对外披露了AI业务的详细收入。这在国内外互联网大厂中都是史无前例的表现。 百度AI业务第三季度收入同比增长超50%，其具体业务包括AI云、AI应用和AI原生在线营销服务等。当下，百度的AI战略正在从投入期转向收获期，AI 公有云连续六年保持市场第一，AI业务收入接近百亿大关，萝卜快跑成为全球自动驾驶领导者，这都证明了百度AI转型的正确性。 结合财报数据可以看出，百度正在更加激进地向AI时代冲刺。对此，主流券商机构普遍持看好态度，他们认为，百度凭借深厚的技术积累和完整的生态 布局，已经度过了漫长的投入期，正步入商业化价值加速释放的新阶段。随着AI应用临界点的到来，百度的先发优势和全栈能力有望转化为持续的业绩 增长和市场份额提升。 AI收入近百亿 百度AI收入主要来自AI云、AI应用和AI原生在线营销服务。其中，AI云收入同比增长33%；AI应用收入达26亿元；AI原生营销服务收入同比增长262%达 28亿元。 云收入是百度AI收入的大头，第三季度AI云收入同比增长33%，其中AI高性能计算基础设施的订阅收入同比增长128%。 云收入主要来自企业客户的采购。根据IDC发布的 ...

数据共享、算法优化和算力协同是"人工智能+"的核心动力。今年5月，之江实验室把太空计算卫星星座 送上太空，通过在轨实时处理数据，有望解决传统卫星数据处理效率瓶颈问题，推动人工智能在太空的 应用发展。"太空中有了计算星座，卫星计算资源就可以快速响应，工作效率明显提升。"在中国工程院 院士、之江实验室主任王坚看来，科技创新发展的每一步都伴随着开放的力量。从过去源代码的开放逐 渐走向数据开放、算力开放，人工智能已经成为一种能够打破学科壁垒的通用语言，而不仅是对科学研 究的简单赋能。 在实数融合大趋势下，"人工智能+"加速推动产业结构系统性优化与整体性升级，更好助力现代化产业 体系建设。百度集团副总裁、深度学习技术及应用国家工程研究中心副主任吴甜认为，深度学习是人工 智能的关键核心技术，驱动人工智能实现了飞跃式进步。百度基于文心大模型与产业级深度学习平台飞 桨打造人工智能基础技术底座，助力千行百业降本增效，提升了相关企业的经营和决策能力。 培育发展战略性新兴产业和未来产业是加快形成新质生产力的重要路径。2025年世界科技与发展论坛近 日在京举行。与会嘉宾一致认为，"十五五"时期将是人工智能技术加速创新、广泛应用的关键 ...

这一次，百度跳出了传统意义上的互联网语境。 百度世界2025大会的聚光灯，投向了一些看起来有些"慢"的领域。相应的，这场大会展示出来一幅更加重资产、长周期的AI图景。 在慧播星数字人、AI搜索、秒哒等一系列AI应用发布了突破性进展的同时，令人眼前一亮的，还有累计服务超1700万次的萝卜快跑无人车，以及同时发 布的新一代国产自研昆仑芯与天池节点。 李彦宏在大会上指出："只有当AI被内化为一种原生的能力，才能真正在各行各业实现效果的涌现。" 这句话也道破了百度"长期主义"的AI战略：百度长达十年的"起大早"，目的从来不是抢占即时风口，也不再只是争夺互联网时代的轻快变现，而是致力于 构建从芯片、框架、模型到应用的全栈能力，将AI锻造成一种深入产业肌理、面向全社会的原生能力。 而今天的智能涌现，比起炫技，更重要的也恰恰是啃下产业深层需求的"硬骨头"。不少声音说百度在AI这件事上"起大早，赶晚集"，但逻辑可能恰恰相 反。 百度看到的，向来不只是AI能生成多么逼真的图片和视频。从算力集群到出行服务，从基础模型到应用落地，正是因为百度很早就开始面向那些真正 的"大集"持续深耕，如今才能手握难以在短期内被复制的核心资产。 看 ...

Core Viewpoint - The article emphasizes the importance of Baidu's long-term strategy in AI development, focusing on building foundational capabilities rather than seeking immediate gains in the fast-paced internet culture [5][12][29]. Group 1: AI Strategy and Development - Baidu's AI strategy is characterized by a commitment to "long-termism," aiming to integrate AI as a fundamental capability across various industries rather than just a tool for quick monetization [5][12]. - The company has made significant investments in foundational technologies, including self-developed chips and AI frameworks, to support its AI applications [14][20]. - Baidu's approach contrasts with the prevailing "speed wins" mentality of the internet era, suggesting that AI competition may not follow the same rules as previous tech races [9][12]. Group 2: Technological Advancements - Baidu has introduced several AI applications, such as the "Luo Bo Kua Pao" autonomous vehicle service, which has completed over 17 million rides, demonstrating the viability of autonomous driving technology [4][18]. - The company has developed the Kunlun chip, capable of supporting multiple large models simultaneously, showcasing its commitment to self-sufficiency in computing power [14][22]. - Baidu's deep learning platform, PaddlePaddle, represents a significant step towards creating a competitive domestic AI framework, challenging the dominance of international platforms [14][22]. Group 3: Market Position and Future Outlook - Baidu's focus on "internalizing AI capabilities" aims to create a sustainable competitive advantage by embedding AI deeply within its business processes [13][24]. - The company is positioning itself as a "military supplier" of AI capabilities, offering modular solutions to businesses, which could lead to widespread adoption of AI across various sectors [24][26]. - Baidu's long-term investments and strategic choices are expected to yield significant returns as the AI landscape matures, moving from a focus on individual breakthroughs to a comprehensive ecosystem of AI applications [28][29].

【环球网科技综合报道】11月3日，在"2025百度十大科技前沿发明"发布会期间，百度公布了涵盖大模型、深度学习框架、AI算力、智能体、AI搜索、数字 人、无人驾驶等AI技术突破，展现AI应用加速走向效果涌现背后的底层创新实力。 当前，国家政策支持AI大模型广泛应用，深入实施"人工智能+"行动方案，人工智能迎来爆发式发展阶段，AI技术加速迭代，应用加速落地效果涌现。百度 持续自主创新，攻关人工智能关键核心技术，积累自主知识产权，以前沿创新引领产业的发展趋势。 会上，百度专利事务部总经理崔玲玲发布"2025百度十大科技前沿发明"并表示，今年发布的百度十大科技前沿发明，涵盖了人工智能算力、框架、模型、场 景应用全栈技术体系。每一项成果，都代表着对技术边界的勇敢突破，彰显着百度在创新道路上的坚定决心与不懈追求。这些前沿技术正汇聚成澎湃的 AI 势能，驱动着生产生活的智能化跃迁。未来，百度将持续围绕关键核心技术攻关，积累自主知识产权，加快专利的产业化进程，赋能千行百业，支撑高水平 科技自立自强和高质量发展。 据介绍，百度发布的2025十大科技前沿发明，包括"自回归统一建模的原生多模态大模型"、 "大模型训练全流程高效容 ...

Core Insights - The event on October 28 marked the launch of AtomGit's upgrade and the AI open-source community, emphasizing the importance of open-source as a foundation for technological competition and industry collaboration [4]. Group 1: Open Source Initiatives - Baidu has led over 1,000 open-source projects, with more than 21,000 contributors in the community [1]. - Since 2009, Baidu has been involved in open-source projects, including the large-scale customization of Hadoop, and has released projects like PaddlePaddle, Apollo, and ECharts [4]. - Baidu's dual-layer open-source model, exemplified by PaddlePaddle and the Wenxin model, has released 10 models in the ERNIE 4.5 series, with parameter scales of 47 billion and 3 billion for mixed expert models, and 0.3 billion for dense parameter models [4]. Group 2: Community Engagement and Impact - As of September 2025, PaddlePaddle has over 90 open-source repositories with a total of more than 220,000 stars, and the Wenxin ecosystem has reached 23.33 million developers and served 760,000 enterprises [5]. - The Xinghe community offers comprehensive services to developers, including 7 million practical projects, over 600 open courses, and more than 400 AI competitions [5]. - Baidu's open-source models have consistently ranked at the top of international model leaderboards, with the latest model, ERNIE-4.5-21B-A3B-Thinking, achieving first place in both global and text model trend rankings [5].

Core Insights - The 2025 World Forum on Science and Technology Development opened in Beijing, highlighting the role of deep learning technology in AI-driven industrial digital transformation [1] - Deep learning is identified as a key technology that has significantly advanced AI capabilities, providing a foundation for the emergence of large models [1][2] Deep Learning Platform - The deep learning platform connects hardware (chips) with large models and applications, essential for AI development, training, inference deployment, and industrial implementation [2] - Baidu's PaddlePaddle serves as an industry-grade open-source deep learning platform, supporting the evolution of the Wenxin large model through a comprehensive framework, model library, and development tools [2][3] - PaddlePaddle has adapted to over 60 chip series and created more than 1.1 million models, showcasing its extensive capabilities [3] Model Performance and Optimization - The collaboration between PaddlePaddle and Wenxin has led to significant performance improvements, including a 47% increase in pre-training MFU for the ERNIE-4.5-300B-A47B model [3] - The model achieves high throughput performance, processing 57K tokens per second for input and 29K tokens per second for output under specific latency conditions [3] Industry Applications - The Wenxin large model has been recognized for its performance in various benchmarks, ranking first in domestic evaluations for multimodal and precise instruction-following tasks [4] - Baidu's deep learning platform is crucial for maximizing the impact of large models across industries, enhancing efficiency and decision-making capabilities [4] Specific Use Cases - In smart manufacturing, CRRC Group utilized PaddlePaddle to reduce high-speed train design simulation time from days to seconds [5] - In smart healthcare, AI optimizes patient experience and doctor efficiency through various stages of medical processes [5] - In smart energy, Baidu's technology has enabled comprehensive monitoring and intelligent decision-making for over 600 plants and 90 sections in the Guangxi power grid [5] Digital Human Technology - Baidu's digital human technology integrates multiple innovative techniques, resulting in highly interactive and realistic digital personas [6] - The commercial value of digital humans is evident, with over 100,000 digital anchors created, achieving a 31% increase in live streaming conversion rates and an 80% reduction in broadcasting costs [6] - The application of digital humans has outperformed real individuals in online performance, setting new records in sales during live broadcasts [6] Developer Engagement - The number of developers using PaddlePaddle and Wenxin has surpassed 23.33 million, serving over 760,000 enterprises [6]

随着人工智能技术迅猛迭代，大模型已成为驱动经济社会发展的关键引擎。在单一模型技术不断突破、 参数纪录屡屡刷新的同时，各家大模型厂商不约而同将目光投向大模型服务链建设，着重构建全场景生 态。 国务院不久前印发的《关于深入实施"人工智能+"行动的意见》提出，发展"模型即服务""智能体即服 务"等，打造人工智能应用服务链。面对千行百业智能化转型需求，全链条覆盖的服务能力、优秀的全 场景生态构建能力，已成为大模型产业下一阶段发展的重点。 加强软硬件协同 对如今的大模型行业来说，"周周有发布，天天有更新"已成为常态。 9月9日，百度正式发布文心大模型X1.1。它采用迭代式混合强化学习训练框架，通过混合强化学习，协 同提升通用任务和智能体任务的效果；同时，通过自蒸馏数据的迭代式生产及训练，不断提升模型整体 效果。尤其在更复杂的长程任务场景中，模型表现突出。 同样在9月，阿里巴巴旗下通义千问也推出了其最新模型，在中英文理解、复杂指令遵循、工具调用等 维度实现性能增强，同时显著减少了知识幻觉。 在追求性能指标不断提升的同时，大模型产业也正走向以配套工具开发、应用场景适配、产业落地为核 心的全场景生态能力建设阶段。 大模型生态 ...

Group 1 - The WAVE Summit 2025 deep learning developer conference was held, coinciding with the launch of China's first national science popularization month [1] - The China Science and Technology Museum and Baidu announced the launch of the first "Science Museum Intelligent Body Matrix" and the "Future Wings Science Popularization Co-Creation Plan" [1][5] - The "Science Museum Intelligent Body Matrix" integrates services from multiple science museums and features various interactive functions, enhancing the immersive experience for visitors [5][6] Group 2 - Over 70,000 visitors have utilized the "Fat Science" intelligent body for exhibit explanations and navigation services [5] - The collaboration includes 27 science museums across China, marking a significant upgrade in science popularization services [5] - The "Future Wings Science Popularization Co-Creation Plan" aims to open data resources and application scenarios, promoting a more open and inclusive science popularization ecosystem [6]