Workflow
多智能体协作
icon
Search documents
假期被玩坏了的奥特曼,正在玩弄全世界的算力
Hu Xiu· 2025-10-07 23:25
如今,春天英伟达的GTC,秋天OpenAI的DevDay,是硅谷最重要的盛会。它们预告了未来。 在刚过去的OpenAI开发者日上,除了发布了ChatGPT Apps SDK、AgentKit、GPT-5 Codex,奥特曼提到的几组截至2025年的数据,揭示AI 行业正在驶向何方: 让我们就此做一点大胆地假设与简单的计算。 第一,OpenAI整体每月tokens消耗。 基于API的调用,显然不是OpenAI对外提供AI服务的全部。去年,OpenAI曾披露它的基于ChatGPT的消费者订阅业务,收入占比约为 75%。而按照OpenAI对未来的收入路线规划,ChatGPT订阅收入占比将逐步下滑,取而代之的是API、Agents与其他新业务占比。当然, 目前OpenAI正在布局Agents相关产品线,并开始探索广告与电商业务,但相比其他两大业务,可以说仍然处于商业化早期阶段。 不妨让我们假定,目前,OpenAI来自API的收入,仍然占据25%,而且,收入占比对应着token的消耗数量。 那么,OpenAI基于API的token消耗量,每周将达到60*60*24*7=60万亿tokens,每月则约为260万亿。相应的 ...
龙湖如何用Agent重塑地产与物业的运营方式?
虎嗅APP· 2025-09-29 00:19
Core Viewpoint - The article emphasizes the transformative role of AI, specifically through the implementation of Agent intelligent systems, in enhancing operational efficiency and decision-making processes within the real estate and property management sectors, particularly for Longfor Group [5][6]. Group 1: AI Implementation and Impact - Longfor Group has integrated AI to address traditional inefficiencies, such as lengthy pricing approval processes that previously took 1-2 weeks, now reduced to a few hours with the use of intelligent agents [5][10]. - The introduction of Agent systems has significantly improved the efficiency of parking lot management, reducing manual audits from 100% to 17%, resulting in an 83% increase in audit efficiency and preventing millions in potential losses annually [15][13]. - The company has developed over 180 digital employees across various scenarios, including pricing models, contract reviews, and risk management, forming a comprehensive digital workforce [6][9]. Group 2: Specific Use Cases - The first successful application of an Agent was in pricing adjustments, where it simulated various pricing scenarios and provided recommendations, thus avoiding potential losses of millions [10][11]. - In property management, the Agent system has automated the identification of abnormal parking lot entries, leading to significant labor cost savings and improved compliance [15][16]. - The contract review process has been enhanced by Agents that automatically identify risks in complex commercial contracts, allowing staff to focus on critical issues [20][17]. Group 3: Challenges and Solutions - The primary challenge in implementing Agent technology is the need for a well-prepared knowledge base within the company, as the complexity of real estate rules and processes requires clear standard operating procedures (SOPs) [21][22]. - Data quality is crucial for the successful deployment of Agents, necessitating a unified digital framework to ensure high-quality data availability [21][22]. - The company has adopted a dual-role team approach, pairing AI product managers with business experts to ensure both technical feasibility and business relevance in project execution [28]. Group 4: Future Directions - Longfor Group plans to expand the use of Agents in internal management, commercial operations, and customer service, focusing on data-driven decision-making and enhanced user experiences [33][34]. - The company is also exploring multi-agent collaboration for complex tasks, demonstrating the potential for creating a virtual employee team that streamlines operations across departments [26][24].
智能体崛起:运营商竞逐下一代数字入口
21世纪经济报道见习记者冉黎黎 北京报道 "现在我们进入智能体的时代,整个发展还有很大的潜力,我们对它充满了期待。"9 月24日—26日,2025年中国国际信息通信展览会(PT展)在北京国家会议中心举行,中国工程院院士邬贺铨在PT展主论坛上如 是表示。热切的期待中,从运营商到产业链企业均在进行智能体的部署。中国联通人工智能首席科学家廉士国预测,未来各个 行业里都会有两类智能体应用,一类是模仿人类经验学习"职业技能",另一类则是"超人专家",可以比人类更早地从复杂的原 始数据里找出规律,带来从0到1的突破。 对于智能体未来将处于日常生活中的何种位置,华为公司ICT Marketing与解决方案销售总裁刘康将之与APP进行了类比。"过去 十年,超过300万款APP深度地融入到我们的生活和工作当中,未来终端会走向以Agent为中心。"刘康表示,"到2030年Agent数 量将超过传统的APP,每个人都会有自己专属的智能助手。" 9月24日—26日,2025年中国国际信息通信展览会(PT展)在北京国家会议中心举行。(冉黎黎/图) 从运营商的角度来说,中国移动集团首席专家、研究院网络与IT技术研究所所长张昊告诉21世纪 ...
思必驰AI办公本X5系列:以多智能体协作与端侧大模型重塑办公效率
Xi Niu Cai Jing· 2025-09-24 09:52
Core Insights - The home appliance industry is entering a critical period of policy effect transition and market demand adjustment in 2025, with overall negative growth becoming a consensus due to the diminishing impact of national subsidies and weak consumer demand [1][6][13] - The promotional rhythm in the industry is tightly connected, with offline channels focusing on the National Day peak season while online platforms prepare for "Double Eleven," leading to differentiated performance across channels [2][10] Policy Impact - The marginal effect of national subsidies is weakening, with retail sales growth for home appliances expected to drop significantly from 23.8% in late 2024 to just 7% by mid-2025 [4][6] - The national subsidy policy has shifted to batch issuance and control, resulting in reduced support for offline channels, which previously benefited from strong subsidy implementation [6][13] Market Performance - The home appliance industry is experiencing negative growth, particularly in traditional categories like refrigerators, washing machines, and air conditioners, with refrigerators expected to see a decline exceeding 20% [6][9] - Online channels are anticipated to outperform offline channels during the promotional periods due to the lower baseline from last year's strong subsidy-driven growth [2][4] Sales Data - For the refrigerator category, online sales volume decreased by 23.8% year-on-year, while offline sales dropped by 20.3%, indicating a significant overall decline in the market [7][9] - Air conditioning sales are projected to decline by 8% in volume and 14.4% in revenue during the "Double Eleven" period, reflecting the ongoing price war and market challenges [8][9] Strategic Recommendations - Companies are advised to focus on retail-driven strategies to accelerate inventory turnover and optimize cash flow, shifting from channel-centric to end-user retail thinking [14] - Emphasis on product structure improvement is recommended to counteract the decline in subsidies by promoting higher value-added products [14] - The industry should leverage the upcoming energy efficiency standard upgrades as an opportunity to launch new products and capture market share [14]
AI办公本是如何弯道超车的?
虎嗅APP· 2025-09-24 09:37
商业世界里,那些有关"后来者居上"的故事,总是为人们津津乐道。 在所有人认为本地生活格局尘埃落定时,没有人能料到阿里、京东会撕开一道口子。当新茶饮行业逼 近天花板时,弯道超车的霸王茶姬凭借大单品模式创造营收和增速的奇迹。同样,在智能硬件这个巨 头盘踞,鲜少能有新鲜事的赛道,谁能想到在2024年还会有新鲜品类能撕开缺口? 思必驰就是这样一位标准的"后来者"。在语音技术领域做了17年B端生意的它,突然在2024年的夏 天,发售了首款AI笔记本,正式进军智能办公的C端市场。 这 场 本 没 有 人 看 好 的 跨 界 冒 险 , 在 短 短 两 年 内 迎 来 反 转 —— 今 年 618 期 间 , 思 必 驰 AI 办 公 本 Pro×LAMY凌美联名款强势拿下京东、抖音双平台彩屏电纸书销售额冠军。2025年9月24日发售的 X5系列产品更成为业内首款实现"彩屏+端侧大模型"的AI办公本。 在X5正式发售前,虎嗅与思必驰IOT事业部首席产品官马斌斌展开了一场对话,试图还原他们如何 从备受质疑的跨界新人成长为逆势突围的行业黑马—— 不重复先行者的老路,而是走出了独属于自 己的"第二条路",这或许是思必驰逆袭的秘密 ...
苦战七年卷了三代!关于BEV的演进之路:哈工大&清华最新综述
自动驾驶之心· 2025-09-17 23:33
Core Viewpoint - The article discusses the evolution of Bird's Eye View (BEV) perception as a foundational technology for autonomous driving, highlighting its importance in ensuring safety and reliability in complex driving environments [2][4]. Group 1: Essence of BEV Perception - BEV perception is an efficient spatial representation paradigm that projects heterogeneous data from various sensors (like cameras, LiDAR, and radar) into a unified BEV coordinate system, facilitating a consistent structured spatial semantic map [6][12]. - This top-down view significantly reduces the complexity of multi-view and multi-modal data fusion, aiding in the accurate perception and understanding of spatial relationships between objects [6][12]. Group 2: Importance of BEV Perception - With a unified and interpretable spatial representation, BEV perception serves as an ideal foundation for multi-modal fusion and multi-agent collaborative perception in autonomous driving [8][12]. - The integration of heterogeneous sensor data into a common BEV plane allows for seamless alignment and integration, enhancing the efficiency of information sharing between vehicles and infrastructure [8][12]. Group 3: Implementation of BEV Perception - The evolution of safety-oriented BEV perception (SafeBEV) is categorized into three main stages: SafeBEV 1.0 (single-modal vehicle perception), SafeBEV 2.0 (multi-modal vehicle perception), and SafeBEV 3.0 (multi-agent collaborative perception) [12][17]. - Each stage represents advancements in technology and features, addressing the increasing complexity of dynamic traffic scenarios [12][17]. Group 4: SafeBEV 1.0 - Single-Modal Vehicle Perception - This stage utilizes a single sensor (like a camera or LiDAR) for BEV scene understanding, with methods evolving from homography transformations to data-driven BEV modeling [13][19]. - The performance of camera-based methods is sensitive to lighting changes and occlusions, while LiDAR methods face challenges with point cloud sparsity and performance degradation in adverse weather [19][41]. Group 5: SafeBEV 2.0 - Multi-Modal Vehicle Perception - Multi-modal BEV perception integrates data from cameras, LiDAR, and radar to enhance performance and robustness in challenging conditions [42][45]. - Fusion strategies are categorized into five types, including camera-radar, camera-LiDAR, radar-LiDAR, camera-LiDAR-radar, and temporal fusion, each leveraging the complementary characteristics of different sensors [42][45]. Group 6: SafeBEV 3.0 - Multi-Agent Collaborative Perception - The development of Vehicle-to-Everything (V2X) technology enables autonomous vehicles to exchange information and perform joint reasoning, overcoming the limitations of single-agent perception [15][16]. - Collaborative perception aggregates multi-source sensor data in a unified BEV space, facilitating global environmental modeling and enhancing safety navigation in dynamic traffic [15][16]. Group 7: Challenges and Future Directions - The article identifies key challenges in open-world scenarios, such as open-set recognition, large-scale unlabeled data, sensor performance degradation, and communication delays among agents [17]. - Future research directions include the integration of BEV perception with end-to-end autonomous driving systems, embodied intelligence, and large language models [17].
外滩大会速递(1):萨顿提出AI发展新范式,强化学习与多智能体协作成关键
Investment Rating - The report does not explicitly provide an investment rating for the industry or specific companies within it. Core Insights - Richard Sutton proposes that we are entering an "Era of Experience" characterized by autonomous interaction and environmental feedback, emphasizing the need for systems that can create new knowledge through direct interaction with their environments [1][8] - Sutton argues that public fears regarding AI, such as bias and unemployment, are overstated, and that multi-agent cooperation can lead to win-win outcomes [9] - The report highlights the importance of continual learning and meta-learning as key areas for unlocking the potential of reinforcement learning [3][13] Summary by Sections Event - Sutton's presentation at the 2025 INCLUSION Conference outlines a shift from static knowledge transfer to dynamic agent-environment interactions, marking a transition to an "Era of Experience" [1][8] - He identifies reinforcement learning as crucial for this transition, but notes that its full potential is contingent on advancements in continual and meta-learning [1][8] Commentary - The report discusses the shift from "data as experience" to "capability as interaction," suggesting that firms need to develop systems that can actively engage with their environments to generate new knowledge [2][11] - It emphasizes that the real bottleneck in reinforcement learning is not model parameters but the ability to handle time and task sequences, highlighting the need for continual and meta-learning capabilities [3][13] Technical Bottlenecks - The report identifies two main constraints in reinforcement learning: the need for continual learning to avoid catastrophic forgetting and the need for meta-learning to enable rapid adaptation across tasks [3][13] - It suggests that R&D should focus on long-horizon evaluation and the integration of memory mechanisms and planning architectures [3][13] Decentralized Collaboration - The report posits that decentralized collaboration is not only a technical choice but also a governance issue, requiring clear incentives and transparent protocols to function effectively [4][12] - It outlines three foundational institutional requirements for effective decentralized collaboration: open interfaces, cooperation-competition testbeds, and auditability [4][12] Replacement Dynamics - Sutton's view on "replacement" suggests that it will occur at the task level rather than entire job roles, urging organizations to proactively deconstruct tasks and redesign processes for human-AI collaboration [5][15] - The report recommends establishing a human-AI division of labor and reforming performance metrics to focus on collaborative efficiency [5][15]
“巨硬”真的来了!马斯克硬刚微软,官宣新公司:要靠 AI “复刻”整个微软
程序员的那些事· 2025-09-11 00:19
Core Viewpoint - Elon Musk's announcement of a new AI software company named Macrohard aims to challenge Microsoft by leveraging AI agents to replicate Microsoft's software capabilities [1][4][12] Group 1: Company Overview - Macrohard is positioned as a purely AI-driven software company, intending to simulate the operations of Microsoft without the need for hardware production [5][6] - The name "Macrohard" was initially a joke made by Musk in 2021, but it has now been formalized into a legitimate business venture [2][4] Group 2: Business Model and Strategy - The core logic behind Macrohard is that AI can perform the same functions as a traditional software company like Microsoft, focusing on software products and subscription services [5][6] - Macrohard will utilize a multi-agent system where hundreds of specialized AI agents will collaborate on tasks such as programming, image/video generation, and user interaction simulations [6][7][8] Group 3: Technological Infrastructure - The backbone of Macrohard's operations is expected to be supported by the Colossus 2 supercomputer cluster, which is being developed by xAI and will feature 1 million NVIDIA GPUs, significantly enhancing computational power [9][10] - Colossus 2 is projected to achieve peak performance between 2000-4000 EFLOPS, marking a fivefold increase from the current Colossus setup [10] Group 4: Competitive Landscape - Microsoft has been a significant player in the AI space, investing over $10 billion in OpenAI and integrating AI models into its products [11] - Musk's criticism of OpenAI and its partnership with Microsoft highlights a competitive tension, with Macrohard representing a direct challenge to Microsoft's dominance in the software industry [11][12]
多智能体的协作悖论
3 6 Ke· 2025-08-27 13:44
Core Viewpoint - The article discusses the emerging trend of collaborative AI systems, where multiple AI agents work together like a human team, potentially surpassing the limitations of single large models [1][2]. Group 1: Collaborative AI Systems - According to IDC, by 2027, 60% of large enterprises are expected to adopt collaborative AI systems, improving business process efficiency by over 50% [2]. - Collaborative AI systems consist of multiple autonomous agents that can perceive, decide, act, and communicate with each other, leading to enhanced problem-solving capabilities [4]. - The performance of multi-agent systems can exceed that of the best single agent by significant margins, as demonstrated by the Claude Opus system, which outperformed the strongest single agent by 90.2% without a substantial increase in generation time [5]. Group 2: Advantages and Challenges - Multi-agent collaboration allows for parallel processing of tasks, significantly reducing task completion time without sacrificing efficiency [5]. - However, the complexity of coordination increases with the number of agents, leading to potential miscommunication and decreased accuracy in outputs [6][8]. - High communication costs can lead to increased computational resource consumption, with token usage in multi-agent interactions being up to 15 times higher than standard conversations [8]. Group 3: Management and Coordination - To manage the complexities of multi-agent systems, a coordinator agent can be introduced to oversee task distribution and conflict resolution, ensuring alignment towards common goals [10]. - Standardized communication protocols can help reduce integration complexity and facilitate efficient information exchange among agents [13]. - The balance between distributed decision-making and centralized control is crucial for the effective functioning of multi-agent systems, requiring ongoing advancements in technology for reliability and security [14].
最新智能体自动操作手机电脑,10个榜单开源SOTA全拿下|通义实验室
量子位· 2025-08-25 23:05
Core Viewpoint - The article discusses the launch of the Mobile-Agent-v3 framework by Tongyi Lab, which achieves state-of-the-art (SOTA) performance in automating tasks on mobile and desktop platforms, showcasing its ability to perform complex tasks through a multi-agent system [2][9]. Group 1: Framework and Capabilities - The Mobile-Agent-v3 framework can independently execute complex tasks with a single command and seamlessly switch roles within a multi-agent framework [3][9]. - It has achieved SOTA performance across ten major GUI benchmarks, demonstrating both foundational capabilities and reasoning generalization [9][11]. Group 2: Data Production and Model Training - The framework relies on a robust cloud infrastructure built on Alibaba Cloud, enabling large-scale parallel task execution and data collection [11][13]. - A self-evolving data production chain automates data collection and model optimization, creating a feedback loop for continuous improvement [13][15]. - The model is trained using high-quality trajectory data, which is generated through a combination of historical task data and large-scale pre-trained language models [22][23]. Group 3: Task Execution and Understanding - The framework emphasizes precise interface element localization, allowing the AI to understand the graphical interface effectively [18][19]. - It incorporates complex task planning, enabling the AI to strategize before executing tasks, enhancing its ability to handle long-term and cross-application tasks [21][22]. - The model understands the causal relationship between actions and interface changes, which is crucial for effective task execution [24][25]. Group 4: Reinforcement Learning and Performance - The Mobile-Agent team employs reinforcement learning (RL) to enhance the model's decision-making capabilities through real-time interactions [28][29]. - An innovative TRPO algorithm addresses the challenges of sparse and delayed reward signals in GUI tasks, significantly improving learning efficiency [31][36]. - The framework has shown a performance increase of nearly 8 percentage points in dynamic environments, indicating its self-evolution potential [36][40]. Group 5: Multi-Agent Collaboration - The Mobile-Agent-v3 framework supports multi-agent collaboration, allowing different agents to handle various aspects of task execution, planning, reflection, and memory [33][34]. - This collaborative approach creates a closed-loop enhancement pipeline, improving the overall efficiency and effectiveness of task execution [34][35]. - The framework's design enables AI to act with purpose, adjust based on feedback, and retain critical information for future tasks [35][36].