机器之心报道 机器之心编辑部 研究多智能体必读指南。 「Anthropic 发布了他们如何使用多个 Claude AI 智能体构建多智能体研究系统的精彩解释。对于任何构建多智能体系统的人来说，这是一本必读的指南。」刚 刚，X 知名博主 Rohan Paul 强力推荐了 Anthropic 一项新研究。 最近一段时间，关于智能体的研究层出不穷。但这也为广大研究者带来一些困惑，比如什么任务需要多智能体？多个 AI 智能体如何协作？怎么解决上下文和记忆 问题…… 面对这些问题，你不妨读读 Anthropic 的这篇文章，或许能找到答案。 文章地址：https://www.anthropic.com/engineering/built-multi-agent-research-system 多智能体系统的优势 有些研究涉及开放式问题，这类问题往往难以预先确定所需的步骤。对于复杂问题的探索，人类无法硬性规定固定路径，因为这一过程本质上是动态且具有路径 依赖性的。当人们开展研究时，通常会根据发现持续调整方法，沿着调查过程中浮现的线索不断推进。 这种不可预测性使得 AI 智能体特别适合执行研究类任务。研究工作要求具备灵活性 ...

2025年6月13日，Anthropic工程团队发布了一份详尽的技术文档，公开了他们构建多智能体研究系统的完整历程。这个系统让多个Claude智能体协同工作，能够搜索 网络、Google Workspace以及各种集成系统，完成复杂的研究任务。 在这份万字长文中，工程团队分享了一个关键数据：以Claude Opus 4为主智能体、Claude Sonnet 4为子智能体的多智能体系统，比单独使用Claude Opus 4的性能提 升了90.2%。但这种性能提升是有代价的——多智能体系统使用的token是普通聊天的15倍。 文章揭示了多智能体系统的核心架构：一个主智能体负责分析用户需求、制定研究策略，然后创建多个子智能体并行探索不同维度的信息。每个子智能体拥有独立 的上下文窗口，能够同时处理不同的搜索任务，最后将结果汇总给主智能体。这种设计让系统能够在几分钟内完成原本需要几小时的研究工作。 在构建这个复杂系统的过程中，团队总结出了8条提示工程原则： 在构建从原型到生产级产品的过程中，Anthropic团队分享了关键的工程实践： 资源分配需要明确规则。 早期版本曾为简单查询生成50个子智能体，陷入失控。解决方案是 ...

图片来源： Sequoia Capital 当算力、模型、分发路径与用户习惯同时成熟， AI 应用的黄金时代悄然来临。真正的主战场，已经从 " 训练最强模型 " 转向 " 谁能让 AI 真正用起来 " 。在红杉 资本的这场闭门分享会上， Pat Grady 、 Sonia Verma 和 Constantine Vassilev 共同抛出四个关键问题 ——What is it 、 So what 、 Why now 、 Now what ？ —— 试 图为这场史无前例的智能浪潮划定航向。 他们给出的判断令人震撼： AI 不仅正在颠覆服务业，还将在十年内撕裂整个软件产业的利润结构 。传统的工具型企业将被 AI 重构为结果导向型组织 —— 不 是 " 卖软件 " ，而是 " 卖结果 " ；不是 " 雇佣人 " ，而是 " 调用 Agent" 。 更具争议的是红杉的核心观点之一： AI 创业的决定性战役将发生在应用层 ，而非底层模型。这与过往 " 大模型赢家通吃 " 的逻辑背道而驰，却也解释了为何 OpenEvidence 、 Glean 、 Harvey 等应用型 AI 公司迅速崛起。 " 智能体即公司，应用层即 ...

Core Insights - The impact of generative AI on enterprises is still limited, with significant changes in specific fields like image generation, translation, and coding, but not in overall business processes and efficiency [3][4] - Many enterprises need to enhance their digitalization before fully leveraging AI technologies, as a solid information technology foundation is crucial for AI applications [4][5] - Companies often misunderstand AI's capabilities, believing it to be omnipotent and easy to implement, which leads to miscalculations in return on investment (ROI) [5][6] Digitalization and AI Investment - Enterprises must develop knowledge bases and intelligent Q&A systems as foundational AI investments, which are essential for future growth [7] - AI investment should be viewed in three areas: hardware, software platforms, and personnel training, all of which are necessary for successful AI implementation [6][8] - The current AI investment landscape is characterized by a focus on future potential rather than immediate financial returns, with many companies still in the early stages of development [8][9] Market Dynamics and Strategy - The AI market consists of three main types of companies: those with large models and cloud services, those offering customized development, and those like IBM that provide integrated platforms [9][10] - IBM's hybrid cloud and AI strategy, established seven years ago, positions the company well in the market, leveraging a wide range of software tools for client needs [10] - The focus of IBM's services is shifting towards private and overseas enterprises, building on its long-standing relationships and trust within the Chinese market [10]

Core Viewpoint - The article discusses the vision of Chang Gaowei, who aims to create a protocol for communication and collaboration between AI agents, potentially disrupting existing platforms like WeChat and Taobao [3][4][5]. Group 1: Vision and Concept - Chang envisions a future where the internet is fully open and interconnected, allowing direct communication between consumers and service providers without intermediary platforms [3]. - He aims to establish the Agent Network Protocol (ANP), likening it to the HTTP protocol that underpins the internet [6][18]. - The emergence of AI agents presents a new opportunity for an open internet, where personal data sovereignty gains value [20][21]. Group 2: Personal Background and Motivation - Chang has a background in telecommunications and internet industries, having worked at Huawei, a small telecom company, and Alibaba [10][11]. - His decision to leave Alibaba was driven by a strong sense of mission to address the lack of interoperability among major platforms [14][17]. - Despite initial self-doubt and external skepticism, he felt compelled to pursue the development of ANP [9][25]. Group 3: Development and Challenges - The initial months after leaving his job were marked by isolation and uncertainty, as he hesitated to share his project due to its ambitious nature [25][27]. - Positive feedback began to emerge after he shared his project within tech communities, leading to connections with supportive individuals [28][31]. - The protocol's first version faced challenges, but external validation and interest from the industry began to grow [36][43]. Group 4: Industry Context and Opportunities - The rise of AI protocols, such as the Model Context Protocol (MCP) by Anthropic, has increased interest in protocol development [36][37]. - Chang identifies a temporary window of opportunity to unite non-platform entities, including startups and hardware manufacturers, under a neutral and open community [49][50]. - The competitive landscape is influenced by geopolitical narratives, with Chinese companies showing interest in developing their own protocols [51][52]. Group 5: Community Building and Outreach - Chang has focused on building a community around ANP, with over 200 contributors on GitHub and active discussions among more than 1,100 participants [43]. - He actively engages in outreach, sharing his vision and garnering support from like-minded individuals [54][56]. - The community's growth reflects a broader desire for an open internet, with many individuals resonating with Chang's ideals [56][60]. Group 6: Future Outlook and Philosophy - Chang's perspective on success has evolved, recognizing that the ultimate goal is to contribute to an open internet, regardless of whether ANP becomes the dominant protocol [65][66]. - He remains optimistic about the potential for collaboration and the impact of shared ideals on the future of the internet [66][68].

Core Insights - The era of AI agents is emerging, with a significant shift in technology and business models driven by these agents, surpassing even the popularity of large models [1][9] - HuiZhi Intelligence has positioned itself as a pioneer in the AI agent field, integrating large models with AI agents since late 2022, and emphasizing the necessity of AI agents for everyone in the near future [1][3] Industry Developments - HuiZhi Intelligence has transformed from a "pioneer" to a "leader" in just two years, focusing on real-world applications rather than laboratory parameters [3] - The company is launching the Agent Cloud Intelligent Agent Service Platform in February 2024, which aims to redefine the implementation of AI agents by addressing the customization and iteration challenges faced by businesses [5] Ecosystem Initiatives - To further integrate AI technology with industry scenarios, HuiZhi Intelligence is initiating the AI Agent Master Plan in June 2025, with an investment of over 10 million yuan to build an open and collaborative ecosystem [7] - The plan aims to empower various industry practitioners and create a comprehensive AI agent ecosystem that focuses on value creation and innovation [7][9] Market Strategy - HuiZhi Intelligence plans to incubate 100 benchmark scenarios and cultivate thousands of master teams in the second half of this year, aiming to establish a network of AI applications covering millions of enterprises [7] - The company emphasizes a collaborative approach, inviting entrepreneurs, developers, and industry professionals to join the AI Agent Master Plan to accelerate the commercialization of AI technology [9]

Group 1 - The fourth Baidu Love Procurement Smart Conference focused on the integration of AI technology into the core aspects of the B2B industry, showcasing innovative solutions such as the B2B industry intelligent agent solution and AI empowerment plans [1][18] - The conference attracted over a thousand enterprise representatives and over a million online viewers, marking a significant milestone in the intelligent upgrade of the B2B industry [1][18] - Baidu's Love Procurement aims to redefine online operations in the B2B sector through customized content production and comprehensive online service capabilities, addressing common operational challenges faced by businesses [7][8] Group 2 - The intelligent agent, as highlighted by Baidu's industry search and intelligent agent business manager, is set to become the mainstream form of AI application in the B2B sector, enhancing operational efficiency and reducing costs for businesses [3][8] - The conference introduced the first practical guide for the B2B industry, titled "7 Days to Effectively Utilize AI," which outlines actionable steps for businesses to leverage AI tools [8][10] - The "Super Beyond Plan" was launched to assist enterprises in transitioning from participants to industry leaders, emphasizing content co-creation and resource investment to enhance brand storytelling [14][18] Group 3 - The event featured a marketplace for quality products, allowing manufacturers to showcase their offerings and connect with potential buyers, thus expanding the scope of Love Procurement beyond traditional B2B interactions [16][18] - The integration of AI in procurement processes is transforming decision-making, with over 80% of state-owned enterprises incorporating online procurement as a primary channel [10][12] - Companies utilizing the B2B industry intelligent agent have reported significant improvements in marketing capabilities, with one company experiencing a 36.4% increase in exposure and a 55.6% growth in lead generation [8][12]

夏天与啤酒烤串是绝配,也与绿茵场上激情的呐喊血脉相连。四年磨一剑的FIFA世俱杯战鼓即将擂响,全 球亿万球迷的炽热目光再度聚焦。即使无法亲临纽约赛场,只需在24℃清凉客厅中点亮海信100吋大 屏,100%的赛场震撼即刻呈现。毕竟,客厅里的海信100吋,不仅是100吋大屏世界销量第一,也是已进驻 FIFA国际转播中心的"官方认证"! 在2025FIFA世俱杯赛事期间,海信电视作为2025世俱杯全球官方合作伙伴,其100吋旗舰大屏以及多款产 品入驻FIFA迈阿密总部国际转播中心,支撑赛事筹备。这不仅是海信显示技术实力的又一次高水平国际 认证,更标志着中国科技力量在全球体育产业核心地带的又一次高光突破。 而AI体育智能体的助攻,则让看球体验"开外挂"。只需轻松语音交互,其就能专业解答赛事规则、智能推 荐赛程、实时分析赛事、预测比赛结果等,甚至可以一键识别场上巨星,带孩子恶补球员知识。就像有个 专业足球解说员在旁陪看,家庭客厅宛如专业观赛直播间。 对球迷而言,最极致的幸福莫过于亲自目睹赛场风云。作为100吋大屏市场销量世界第一的海信,其100吋 大屏则能以颠覆性的画质体验,让家庭客厅一秒升级为球场前排VIP席位。 拿 ...

机器之心报道 编辑：泽南、Panda 越通用，就越World Models。 我们知道，大模型技术爆发的原点可能在谷歌一篇名为《Attention is All You Need》的论文上。 如今，在通用人工智能（AGI）初现曙光，Scaling Laws 却疑似接近极限的当口，又是谷歌对未来方向进行了一番思考，想来想去还是只有五个词： 论文标题：General agents need world models 论文链接：https://arxiv.org/abs/2506.01622 香港中文大学博士 Richard C. Suwandi 撰写了一篇博客文章，详细解读了这篇开创性的论文及其对 AGI 未来的意义。 | AUTHORS | AFFILIATIONS | PUBLISHED | | --- | --- | --- | | Richard Cornelius Suwandi | The Chinese University of Hong Kong, Shenzhen | June 11, 2025 | 想象一下，如果我们能够构建一个像人类一样思考和计划的人工智能，未来会是什么样子。大语言模型（L ...

Group 1 - The year 2025 is widely regarded as the "Year of AI Agents" in the global technology and investment markets, with AI agents and their ecosystems seen as significant investment opportunities in the computer sector [1] - The performance gap between different AI models is rapidly narrowing, and domestic computing power supply is maturing, driving demand from traditional industries for AI technology to enhance efficiency and revenue [1][5] - A wave of technological innovation and commercialization surrounding AI agents is sweeping through the global tech sector, with major companies competing to seize the technological and market advantages [1][5] Group 2 - AI agents have evolved from single-task "Chatbots" to "Copilots" and now to "semi-autonomous entities," gaining identity and intent expression capabilities [2] - The new architecture of the AI economy is expected to lead to a revolutionary scenario of "AI hiring AI," where task-leading AI delegates sub-tasks to other AIs, creating a dynamic self-organizing task network [2] - Future development teams may consist of "AI project managers" directing "AI programmers," "testers," and "analysts" to collaborate seamlessly, marking a new dimension of competition for enterprises [2] Group 3 - Major tech companies are launching new competitive initiatives centered around AI agents, with Microsoft and Google leading the charge in creating new application ecosystems [5] - In China, companies are shifting focus from showcasing models to demonstrating the capabilities of AI agents, with a surge of vertical AI products expected in the coming months [5][6] - The launch of the "Nano AI Super Search" by 360 marks a significant milestone, as it is the first domestic product based on super-agent technology, aiming to redefine "AI search" with eight core capabilities [6][12] Group 4 - The "Nano AI Super Search" has quickly gained traction since its launch, becoming one of the largest AI-native search engines globally, ranking fourth in monthly active users in the Chinese market [7][10] - It ranks sixth in the global AI product rankings, being the only domestic product besides DeepSeek to make it into the top ten [10] - The eight core capabilities of the Nano AI Super Search include deep search, complex problem-solving, task decomposition, and multi-model collaboration, enhancing its decision-making, execution, and collaboration abilities [13][14] Group 5 - The Nano AI Super Search aims to realize the vision of "super individuals" and "super companies," where one employee can manage multiple AI agents [12] - It can seamlessly integrate various input methods and emerging hardware, expanding the potential applications of AI agents [15] - The launch of the Nano AI Super Search signifies a shift from AI agent concepts to practical changes in user interaction, strengthening 360's competitive position in the AI product matrix [16]