7月26日，2025世界人工智能大会（WAIC 2025）在上海拉开帷幕。联想集团副总裁、联想中国技术管 理委员会执行主席阿不力克木·阿不力米提（以下简称"阿木"）在当天下午举办的人工智能终端产业发 展论坛上发表主题演讲表示，人工智能终端是终端产业的第三次代际升级，个人超级智能体将成为原生 人机交互系统，混合异构AI计算成为内生标配，"算力+模型能力"的双螺旋进化推动着端侧智能指数级 跃升。 （联想集团副总裁、联想中国技术管理委员会执行主席阿木，图片来源：企业供图） 作为全球和中国终端的引领者，联想正通过"一体多端"的策略实现领跑：不仅率先上市了全球第一台具 备五大特征、真正意义上的AI PC，而且发布了业内首款个人超级智能体——联想天禧，其内嵌于联想 AI PC、AI手机、AI平板和AIoT等"多端"，为用户打造跨平台、跨设备、端云一体、开放互联的无缝融 合AI体验。 "联想坚持AI终端原生技术首创，持续助力产业高质量发展。"面对当前人工智能终端产业发展机遇，阿 木呼吁共建更加繁荣的产业生态，以开放协作和创新发展，共筑中国人工智能终端产业全球竞争力。 AI终端重塑人机交互范式 从功能执行到"硅基分身" 当前 ...

Core Insights - The 2025 Global Digital Economy Conference opened in Beijing, featuring the expansion of the "Human-Centric AI Development and Governance Initiative," which now includes nearly 50 member organizations from various sectors of AI technology and applications [2] Group 1: Initiative and Membership - The "Human-Centric AI Initiative" was jointly launched by several organizations, including Caixin Think Tank and Lenovo Group, in November last year, focusing on technology research, application implementation, and ethical governance [2] - New members of the initiative include major tech companies such as Baidu, MiniMax, Inspur, and Siemens (China), covering key areas like large models, embodied intelligence, and AI healthcare [2] Group 2: AI Development Perspectives - Lenovo's Vice President emphasized the need for hybrid AI development to empower various industries and enhance global competitiveness in the digital economy [3] - The development of hybrid enterprise intelligence relies on upgrading intelligent computing infrastructure and fostering an open application ecosystem [4] Group 3: AI Applications and Innovations - Lenovo has launched personal and enterprise super intelligent agents, such as the Lenovo Tianxi personal super intelligent agent and the "Lenovo Enjoy" enterprise super intelligent agent, aimed at improving quality of life and productivity [4] - The company is also working on city-level super intelligent agents for smart governance in locations like Yichang and Wuyishan [4] Group 4: Standards and International Initiatives - Lenovo has contributed to the publication of the report "Human-Centric AI: Technology Development Perspectives in the Era of Human-Machine Symbiosis" and has participated in establishing international AI standards [5] - In the 2024/25 fiscal year, Lenovo signed three major international initiatives, including the EU's "AI Pact" and UNESCO's ethical AI statement [5]

图①：敬毅民在研究联想多智能体间的任务规划。 殷晨鑫摄 图②：连涛向学生介绍年度科研亮点成果。 胡睿坤摄 图③：靳志文在实验室调试X射线成像系统的光路组件。 本报记者 赵帅杰摄 图④：吴杰（左）在风洞实验室指导研究生。 卫明远摄 又是一年毕业季，新一批青年人即将走上科研岗位，以青春的力量投身科技强国建设，书写各自的奋斗 故事。 科研之路道阻且长，如何选对方向？如何坚定信念走得更好、走得更远？我们邀请了几名在不同领域从 事科研工作的青年科技工作者，请他们讲述自己当初为何选择投身科研，并勇毅前行、不断攀登科技高 峰的故事。 ——编 者 选择人工智能领域，是出于热爱和兴趣 敬毅民 我是一名95后，硕士毕业于苏州大学自然语言处理实验室。这个实验室以自然语言理解、中文信息处 理、机器翻译和自然语言认知为主要研究方向。我的求学、工作经历也始终围绕着人工智能展开。 选择人工智能领域，是出于热爱和兴趣。记得在大学读书时，有时看到一篇高质量的人工智能学术论 文，我就会激动到难以入眠。加入联想公司后，我把这份热爱延续到了工作中。当看到自己的数据挖 掘、算法优化工作成果能够落地并转化到产品上时，我会觉得非常有成就感。在科研院所做研 ...

Core Insights - The next wave of AI is focused on generating revenue rather than just providing tools, which is seen as a trillion-dollar opportunity by industry leaders [2] - The transition from large models to intelligent agents marks a new era in AI, emphasizing automation and cash flow generation [2] - Companies' core competitiveness will depend on customized AI applications and quantifiable business outcomes [2][3] Data and Integration - High-quality data is essential for companies to realize the benefits of AI, with data integration being a critical factor [3] - The integration of AI with traditional automation technologies is a key focus for future AI development, particularly in manufacturing [3][4] Intelligent Agents - The demand for intelligent agents is growing, with various companies launching advanced AI models and solutions [6][7] - IBM has introduced a comprehensive enterprise-ready AI agent solution, emphasizing collaboration and integration with existing IT assets [7][8] Application and Use Cases - Intelligent agents are being applied in specific business scenarios, such as customer service and R&D, to enhance efficiency and reduce operational costs [10][11] - Companies are encouraged to start with small, specific use cases to validate ROI before scaling up [12] Market Trends - The sales of AI agents and related products are projected to significantly increase, with estimates suggesting revenues could reach $125 billion by 2029 and $174 billion by 2030 [6] - The competitive landscape is shifting as companies seek to leverage AI agents for greater returns on investment [12]

新华财经上海5月21日电（记者郭慕清）2025年被称为"AI智能体元年"，人工智能技术正以前所未有的 速度重塑各行业格局。人们昔日在《钢铁侠》里看到的场景——托尼·斯塔克与贾维斯（Jarvis）进行交 流，让AI助手控制各种系统并完成复杂任务……正在从科幻照进现实。 最后是市场转化效率有待提升。即便完成产品开发，规模化交付仍依赖多元销售渠道、专业营销策划及 客户触达能力。部分ISV由于缺乏成熟的推广体系和政企服务经验，导致产品市场转化周期延长。 "相较通用大语言模型，行业智能体需要深耕垂直领域。以金融大模型为例，必须聚焦场景、融合数据 与智能，构建AI协同体系。"人工智能企业第四范式旗下的SageOne先知副总裁李铮说。 企业加速构建生态协同体系 面对市场痛点，多家企业正通过构建AI生态破解发展难题。记者调研发现，品牌力、方案力、营销力 与销售力的四维赋能体系，正成为加速智能体落地的关键路径。 尽管一切"看上去很美"，记者近日调研却发现，这并不是一条坦途，仍有许多关口需要跨越，比如，当 下政务、医疗、金融、制造等行业对AI的关注焦点正从技术能力转向部署方式和实际可用性。原本以 云端服务为主的独立软件开发商（I ...

Core Insights - The AI industry is at a critical juncture, facing challenges in scaling applications despite advancements in generative AI technology [1] - Lenovo's CEO Yang Yuanqing has articulated a vision for AI that focuses on a "super intelligent agent" model to facilitate large-scale application deployment [2][3] Group 1: AI Application and Development - The "super intelligent agent" represents an evolution in AI applications, characterized by cross-device perception, multi-modal interaction, and autonomous task decomposition [2] - Lenovo aims to transition from being perceived solely as a hardware vendor to an AI service provider, integrating AI across all business operations [3][6] Group 2: Features and Capabilities of Super Intelligent Agents - The super intelligent agent is designed to move beyond passive assistance, enabling proactive service based on user intent, such as planning a family trip by coordinating various tasks [4] - In enterprise scenarios, Lenovo's super intelligent agent has been integrated into its operations, showcasing capabilities across multiple domains like supply chain and customer service [4] Group 3: AI Infrastructure and Security - Lenovo has developed a "Lenovo Inference Acceleration Engine" to enhance local inference capabilities on PCs, making them comparable to cloud models [4] - Data security and privacy protection are fundamental to the super intelligent agent's functionality, with measures in place to counter threats like Deepfake attacks [5] Group 4: Market Position and Strategy - Lenovo's AI transformation reflects a broader trend among hardware manufacturers to leverage their extensive device ecosystems for AI opportunities [6] - The company has established a global manufacturing system with 33 factories across 10 countries, allowing for rapid adjustments to market changes and tariff impacts [7][8]

在AI技术快速落地的过程中，数据安全与隐私保护成为杨元庆反复强调的底线。他明确表示："用户隐 私与企业数据安全是联想AI创新的基本原则，不可妥协。保护个人隐私、保护企业知识产权是现在推 广个人智能以及企业智能的主要出发点。" 新华财经上海5月8日电（记者郭慕清、沈寅飞）近日，在联想Tech World 2025创新科技大会上，联想集 团董事长兼CEO杨元庆正式发布覆盖全场景的超级智能体矩阵，包括联想天禧个人超级智能体、联想乐 享企业超级智能体、联想城市超级智能体及新一代联想推理加速引擎，并阐述其技术升级方向与用户服 务能力。 杨元庆在接受新华财经采访时表示，联想天禧具备三大核心能力：更强的环境感知与交互能力、个性化 认知与决策支持、自主任务规划与持续学习能力。用户可通过语音、文字、设备触控等自然方式与智能 体进行交互。 其中，在个性化服务方面，联想天禧通过长期记忆功能记录用户习惯与偏好。例如，用户规划旅行时， 智能体不仅能推荐符合预算的酒店，还能根据历史行程安排交通、预订餐厅，甚至细化到提醒用户携带 特定物品。智能体不再依附于单一设备，而是与"人"绑定，整合用户所有设备的数据，在安全和隐私的 前提下，成为个人 ...

更多AIGC资讯 …… 1.【消息称OpenAI计划削减与微软的收入分成】据报道，OpenAI已经告诉投资者，它将降低与微软 分享的收入比例。该报道称，到本十年末，这一比例将至少下降一半。（新浪财经） 2.【谷歌发布Gemini 2.5 Pro "I/O"模型】当地时间5月6日，谷歌DeepMind人工智能研究团队发布 了Gemini 2.5 Pro "I/O" ，这是今年3月发布的多模态大语言模型Gemini 2.5 Pro的升级版本。在 AI排行榜LMArena上，该模型在多个指标上登顶。目前该模型已向Google AI Studio的独立开发 者、Vertex AI云平台的企业用户以及Gemini应用的普通用户开放。（界面新闻） 3.【星动纪元宣布开源首个AIGC机器人大模型】5月7日，星动纪元宣布开源首个AIGC机器人大模 型。据了解，VPP将视频扩散模型的泛化能力转移到了通用机器人操作策略中，巧妙解决了diffusion 推理速度的问题，让机器人实时进行未来预测和动作执行，大大提升机器人策略泛化性，并且现已全 部开源。（IT之家） 4.【联想发布端侧AI新技术】5月7日，联想创新科技大会（Tech ...

Group 1 - Lenovo officially launched a comprehensive super intelligent matrix at the Tech World 2025 conference, which includes personal, enterprise, and city super intelligent agents, along with a new generation of inference acceleration engines [1] - The exponential growth of edge AI is driven by a dual spiral of computing power and model capability, as explained by Lenovo's CEO Yang Yuanqing [1] - The inference acceleration engine, developed in collaboration with Tsinghua University and Wuneng Chip, addresses industry pain points related to model lightweighting and inference efficiency [1] Group 2 - The core foundation of the enterprise super intelligent agent is based on a computing power support system, utilizing a hybrid computing infrastructure built on a "edge-cloud-network" four-layer architecture [2] - This architecture supports data collection, storage, and processing, enabling the training of enterprise AI models and their deployment on edge devices or terminal hardware for distributed inference [2] - Lenovo's first enterprise super intelligent agent, "Lenovo Lexiang," integrates comprehensive business data and knowledge assets, achieving intelligent solutions across the entire supply chain to marketing [2]

Core Concept - Lenovo's CEO Yang Yuanqing defines "super intelligent agents" as the next breakthrough in enhancing quality of life and productivity, evolving from mere tools to cognitive operating systems for individuals and enterprises [2][5]. Group 1: Definition and Capabilities - Yang outlines three core capabilities of super intelligent agents: perception and interaction, cognition and decision-making, and autonomy and evolution [4][5]. - The super intelligent agent is designed to deeply integrate perception, cognition, decision-making, and autonomous evolution, enhancing user understanding through increased interaction [7]. Group 2: Technological Advancements - The past decade has seen AI assistants evolve from simple voice command execution to multi-modal interaction capabilities, although limitations in computing power and privacy have hindered their full potential [5]. - Lenovo has introduced a reasoning acceleration engine in collaboration with Tsinghua University, aimed at addressing the industry's core challenges of large model lightweighting and inference efficiency [8]. Group 3: Product Launches - At the event, Lenovo unveiled several devices equipped with the Tianxi personal super intelligent agent, including the AI-enabled Moto Razr foldable phone and the ThinkBook with a rollable screen [10]. - The company also introduced the Lenovo Lexiang, the first enterprise super intelligent agent, designed to enhance business operations through its multi-modal capabilities [11]. Group 4: Future Outlook - Yang predicts that the capabilities of edge AI will triple in the next 12 months, driven by advancements in computing power and model capabilities [8]. - The transition from tools to "silicon-based life forms" signifies a shift in how enterprises will leverage super intelligent agents for innovation and operational efficiency [12].