【观网财经丨智能早报 6月12日】 马斯克后悔发布涉特朗普贴文，后者回应 当地时间10日，美国企业家、政府效率部前负责人马斯克在其社交平台表示，他对上周发表的一些关于 美国总统特朗普的帖子感到后悔，称"内容说得太过了"。"我对上周（发布的）关于特朗普总统的一些 帖子感到后悔。太过了。"马斯克在最新帖文中说。 对此，特朗普回应称，"我认为他这样做非常好。"（央视新闻） 钉钉产研负责人离职时发万字长文谈阿里，马云回应 6月11日消息，近日钉钉产研负责人元安（花名）离职时在阿里内网发表了近万字长文，谈及阿里的发 展历程、现存问题及相关建议。阿里创始人马云对该帖作了回复。 元安提到，阿里的使命是让天下没有难做的生意，愿景是成为一家活102年的好公司。自己当初也是受 这一使命感召加入阿里。但他十五年走下来，发现外部社会舆论和阿里内部员工心态，都发生了巨大改 变，已经失去了以前那种要为社会带来美好改变的梦想。 对于元安的分析与建议，马云在阿里内网回复称，"元安同学好，谢谢你那么长的信，写得很好。好像 人的成长，阿里的发展也有很多必然要走的路和过程，阿里巴巴在发生变化之中。祝福你，也希望经常 回来看看。"（智通财经） 喜马 ...

山东大模型产业高质量发展专项指导意见实施一年 芒种时节，位于珠三角的一片水稻试验田里，盘旋的无人机实时采集着株高、籽粒大小、叶片色泽 等水稻表型数据，由此形成的可视化"数字农田"在华南农业大学农学院的实验室里徐徐展开。 这里有一个依托浪潮模块化数据中心解决方案构建的智算中心，华南农业大学由此在推进一场育 种"革命"：传统育种需要育种学家凭借多年经验选取亲本再一代代观察田间表现，如今，大模型帮着科 研人员将每粒种子所包含的"信息天书"快速解码，"算"出具有高产、优质潜力的好种子，育种周期可从 原先的三四十年，缩短至三到五年。 "这一年来模型能力不断'刷新'，有了更多不同参数量级的模型，比如'源2.0'就是国内首个全面开 源的千亿级参数基础大模型。"浪潮集团科技创新部副部长魏磊介绍，浪潮现在有25个大模型，包含1个 基础大模型和24个行业大模型。去年6月，山东行业大模型不足20个，仅有浪潮海若一个通过中央网信 办备案的行业大模型。今年1—4月，全省通过第三方机构测评的大模型产品有204个，11个生成式大模 型通过备案。"上架"的模型在变多，更重要的是模型的场景服务能力迅速成长。"我们的大模型工厂已 能批量满足客户 ...

Group 1: US-China Trade Talks - The first meeting of the US-China trade negotiation mechanism took place in London, led by Chinese Vice Premier He Lifeng and US Treasury Secretary Janet Yellen, among others [2] - Both sides engaged in candid and in-depth discussions, reaching a principled consensus on key economic and trade issues [2] - The meeting aimed to implement the important consensus from the June 5 call between the two countries' leaders and to consolidate the outcomes of the Geneva trade talks [2] Group 2: Automotive Industry Developments - Several automotive companies, including Dongfeng Motor and China FAW, announced a unified payment term for suppliers to within 60 days [3] - This initiative is intended to support the healthy development of small and medium-sized enterprises in line with national policy [3] Group 3: IP Retail Market Growth - China has become the fourth largest IP retail market globally, with an annual retail sales total of 155.09 billion yuan in 2024, reflecting a year-on-year growth of 10.7% [4] - The IP derivative market in China is expected to exceed 100 billion yuan by 2027, with a compound annual growth rate of over 20% [4] Group 4: AI Development Initiatives - Liaoning Province has issued a plan to promote AI innovation, aiming to establish a collaborative development model by 2027 [5] - The plan includes building a computing power infrastructure with a scale of over 5000 PFLOPS and nurturing over 300 AI enterprises [5] Group 5: New Product Launches - Huawei launched the Pura80 series smartphones, featuring HarmonyOS 5.1 and starting prices of 6499 yuan for the Pro model and 9999 yuan for the Ultra model [6] - Alibaba's Qwen3-Embedding model topped the global open-source model rankings, indicating strong interest in AI models [7] Group 6: Corporate Restructuring and Financial Updates - Li Auto announced the establishment of two new robot departments to accelerate its AI strategy [9] - The founder of Pang Donglai projected a net profit of 1.5 billion yuan for the company in 2025, with an average monthly income of 9000 yuan for employees [10] - Marelli Corp filed for Chapter 11 bankruptcy protection in the US and secured $1.1 billion in new financing [14] Group 7: Market and Economic Indicators - The US manufacturing sector is under significant supply pressure due to a shortage of rare earth materials, with current inventories only sufficient for two to three months [11] - Tesla's Robotaxi project is set to launch in Austin, Texas, on June 22, with potential adjustments to the timeline [12] - Qualcomm has established an AI research center in Vietnam, focusing on generative and agent-based AI solutions [13]

Group 1 - Doubao-Seed-1.6 model was officially launched, supporting 256k long context, with a daily average token usage of 16.4 trillion by the end of May [1] - The launch reflects the extensive application of Doubao in a competitive large model industry, necessitating continuous performance optimization and innovation in multimodal fields [1] Group 2 - Qianxun launched a mini-program named "Qianxun Super Member," which sparked speculation about its connection to the return of influencer Viya, but the company denied any such link [2] - The mini-program aims to promote consumption and serve merchants while adhering to national laws and regulations, indicating a shift in the live e-commerce industry towards private domain operations [2] Group 3 - AliExpress, a cross-border e-commerce platform under Alibaba, launched a car sales business, initially offering Chinese electric vehicles during the overseas "618" shopping festival [3] - This move is seen as a strategic expansion into vehicle sales, leveraging its existing user base in auto parts, but faces challenges related to after-sales and logistics [3] Group 4 - 3D model company VAST secured tens of millions of dollars in Pre-A+ round financing, led by Beijing AI Industry Investment Fund, with previous investors including various venture capital firms [4] - VAST launched the world's first AI-driven one-stop 3D workspace, Tripo Studio, and is set to introduce a new algorithm, Tripo 3.0, enhancing its competitive edge in the 3D modeling market [4]

Core Insights - Artificial intelligence (AI) is evolving from a mere tool to an interactive assistant capable of engaging with the real world and humans [1] - The 2025 Zhiyuan Conference highlighted the advancements in AI, focusing on embodied intelligence and general AI, showcasing new ideas and results that facilitate AI's integration into reality [1] Group 1: AI Development and Challenges - The rapid development of generative AI has made large models the core technology supporting AI applications, expanding their capabilities across various scenarios [2] - Experts at the conference emphasized that future AI should not only be able to communicate but also understand, reason, and take action [2] - Concerns were raised about the potential uncontrollable behavior of AI systems, with suggestions for developing "scientist AI" to help humans better understand and analyze the world [2][3] Group 2: Embodied Intelligence - Embodied intelligence emerged as a hot topic at the conference, with demonstrations from various robots showcasing their capabilities [4] - The goal of using AI technology in robotics is to liberate human productivity, with expectations for robots to enter various fields, including healthcare and manufacturing [4] - The upcoming World Humanoid Robot Games aims to connect technology training with real-world applications, addressing genuine enterprise needs [5] Group 3: Technical and Engineering Challenges - Despite the excitement around embodied intelligence, significant technical and engineering challenges remain, impacting the transition from "visually appealing" to "practically useful" robots [6] - Current embodied intelligence models face issues of usability and adaptability, with high difficulty in matching "embodied brains" to actual hardware [6] - The Zhiyuan Research Institute launched the "Wujie" series of embodied intelligence platforms to lower development barriers and foster collaboration within the industry [6][7]

点击按钮▲立即预约 中美贸易磋商原则上达成框架 伦敦时间6月10日，商务部国际贸易谈判代表兼副部长李成钢在伦敦谈到中美经贸磋商机制首次会议时表示，中美双方进行了专业、理性、深 入、坦诚的沟通。双方原则上就落实两国元首6月5日通话共识以及日内瓦会谈共识达成了框架。美国商务部长卢特尼克表示："谈得很顺利，我 们花了很多时间一起交流。" 据悉，在这轮磋商中，美方考虑取消近期出台的一系列限制措施，涉及芯片设计软件、喷气发动机零件、乙烷以及核材料等领域。这些措施大多 是在过去几周中美关系再度紧张升级时出台的，条件是中国放宽对稀土出口的限制。（综合新华社） ｜点评｜ 中美贸易谈判达成初步框架，是双方都乐于见到的结果。中美两国已经用无数次事实证明，两国关系合则两利，斗则两伤。美国经济 增速放缓，劳动力市场降温；中国内需修复缓慢，行业复苏不均衡。中美两国现在都需要一个稳定的外贸环境，稳固国内经济发展局面，而非 给市场带来更多不确定性。 从中美两国历次谈判经验来看，最终两国需要进行多轮谈判，才能恢复到较为常态化的贸易环境中，期间甚至可能发生多次谈判停摆或重启事 件。无论如何，两国贸易代表坐在一张谈判桌上，坦诚地进行交流，远比加 ...

Core Insights - The article discusses the launch of various AI models and services by Volcano Engine at the FORCE Original Power Conference, highlighting advancements in the Doubao model series and the introduction of new video and voice generation models [2][31]. Group 1: Doubao Model Series - The Doubao model 1.6 series includes three models: doubao-seed-1.6, 1.6-thinking, and 1.6-flash, with 1.6-thinking surpassing DeepSeek-R1-0528 in reasoning capabilities [2][8]. - Pricing for Doubao 1.6 is based on input length, with costs significantly reduced to 0.8 yuan per million tokens for input and 8 yuan per million tokens for output, making it one-third the cost of previous models [2][5]. - Doubao model 1.6 supports multi-modal understanding and GUI operations, enabling applications in e-commerce, autonomous driving, and security inspections [10][12]. Group 2: Seedance Video Generation Model - The Seedance 1.0 pro model achieved top rankings in text-to-video and image-to-video tasks, outperforming competitors like Veo3 and Kuaishou 2.0 [5][17]. - The cost of generating a 5-second 1080P video is only 3.67 yuan, allowing for the production of over 2,700 videos within a budget of 10,000 yuan, which is claimed to be the lowest in the industry [5][16]. Group 3: AI Cloud Native Services - Volcano Engine upgraded its AI cloud-native services, introducing tools like MCP service, PromptPilot, and AI knowledge management systems, aimed at enhancing agent development and application [7][21]. - The daily token usage for the Doubao model exceeds 16.4 trillion, reflecting a 137-fold increase since its initial release [7][14]. - The Doubao model holds a 46.4% market share in China's public cloud large model market, ranking first [7]. Group 4: Agent Development Tools - Twelve new agent development tools were introduced, including TRAE for code assistance and MCP service for cloud service integration, which aims to streamline the development process [21][22]. - The PromptPilot application optimizes user prompts and supports complex task management, enhancing the overall efficiency of AI interactions [23][24]. Group 5: AI Infrastructure and Security - Volcano Engine launched a multi-modal data lake solution to enhance data processing and analysis capabilities, significantly reducing data acquisition costs by 80% [26]. - AI security products were introduced, including AICC for secure computation and a large model application firewall to protect against attacks without increasing inference delays [30]. Conclusion - The AI product matrix unveiled by ByteDance at the conference illustrates a comprehensive strategy focused on model capabilities, toolchains, and infrastructure, aiming to enhance industry efficiency and innovation [31][32].

Core Insights - ByteDance's Volcano Engine has launched new AI models, including Doubao Model 1.6 and Seedance 1.0 pro, aiming to enhance AI cloud-native services and support enterprise applications [1][2] - The Doubao model series has seen a significant increase in usage, with daily token usage exceeding 16.4 trillion, a 137-fold increase since its initial release [1] - Doubao Model 1.6 introduces a pricing strategy based on input length, significantly reducing costs for enterprises, making it more accessible for developers and small businesses [2][3] Group 1: Product Launch and Features - Doubao Model 1.6 supports multi-modal understanding and graphical interface operations, enhancing its capability to address real-world problems [1] - Seedance 1.0 pro can generate high-quality 1080P videos with seamless transitions, utilizing text and image inputs [1] - The Doubao model series now encompasses various modalities, including video, image, voice, and music, promoting comprehensive intelligent applications [1] Group 2: Cost Reduction and Market Impact - Doubao 1.6's pricing for the most used input range (0-32K) is set at 0.8 yuan per million tokens for input and 8 yuan per million tokens for output, making it one-third the cost of its predecessor [2] - Seedance 1.0 pro's cost is the lowest in the industry at 0.015 yuan per thousand tokens, with a 5-second 1080P video costing only 3.67 yuan [2] - The price reduction is expected to accelerate technology adoption and lower the barriers for AI transformation in enterprises, benefiting startups and SMEs [2] Group 3: Advancements in AI and Agent Development - The evolution of large models is shifting from perception AI to generative AI and now to Agentic AI, aiming for autonomous reasoning and task execution [3][4] - Volcano Engine has upgraded its AI cloud-native services to support Agent development, including new tools and frameworks [3] - The integration of Doubao 1.6 with ByteDance's AI programming product TRAE has led to over 80% of engineers using it for development, with monthly active users exceeding 1 million [3]

Investment Rating - The report does not explicitly state an investment rating for the GEO industry Core Insights - The GEO industry leverages generative AI technology to create content that aligns closely with user intent, enhancing its ranking and citation in AI searches, emphasizing content interpretability and authority [6] - The market for AI search products shows a significant concentration of traffic among leading players, with DeepSeek and Nano AI dominating the landscape [12][16] - Traditional marketing faces multiple challenges, including trust crises, information gaps, competitive pressure, and content imbalance, which GEO aims to address through targeted solutions [18][28] Summary by Sections GEO Marketing Transformation - GEO utilizes generative AI to optimize content for AI search engines, improving visibility and user engagement [6] - The report outlines the traffic situation for AI search products, indicating a competitive landscape with clear leaders and laggards [9][14] AI Search Product Traffic - In March 2025, DeepSeek led the AI search web traffic with 494.4 million visits, followed by Nano AI with 301.25 million visits, indicating a strong head effect in the market [12] - The application side of AI search shows Quark, Doubao, and DeepSeek as the top three players, with significant user engagement [16] Core Pain Points in Marketing - Companies face trust issues due to exaggerated claims and data privacy concerns, leading to a decline in brand image [24] - Information gaps arise from fragmented content across platforms, making it difficult for users to obtain complete product information [26] - Competitive pressure is evident as leading firms dominate key market segments, making it challenging for newer entrants to gain visibility [27] GEO's Solutions to Marketing Challenges - GEO addresses trust issues by ensuring content accuracy and compliance through advanced technologies [36] - It enhances competitive analysis and strategy formulation to help brands navigate market pressures [29] - GEO promotes user insights by analyzing search behaviors and preferences, aiding in product optimization and content strategy [30] Comparison of Traditional Marketing and GEO - Traditional marketing methods are often costly and slow to yield results, while GEO offers a more efficient, trust-building approach by delivering answers directly to users [38] - GEO's content can be reused across platforms, creating long-term value and reducing marketing costs compared to traditional methods [40]

大模型训推系统宛如一辆精密调校的赛车，即便搭载顶级引擎（高算力芯片），如果油箱（内存）、 变速箱（带宽）与路况（任务类型）不匹配，仍会陷入" 龟速" 困局。华为研究团队发现，超过 60% 的算力浪费在硬件资源错配与系统耦合上，而传统"人拉肩扛"的优化方法在芯片特性的" 三角 矛盾" （算力 - 带宽 - 容量失衡）前束手无策。 大规模训练集群的利用率黑洞 ：大模型训练过程像驾驶中的 " 猛踩油门 " （训练阶段密集计 算）， MoE 模型更如混合动力车，需精准平衡计算与内存配比，极致压缩通信占比，稍有不 慎便效率骤降。 动态实时推理系统任务的两极分化 ：从短问答（城市道路）到长文本生成（越野山路），推理 阶段硬件需同时满足高吞吐与低延迟，如同要求一辆车兼具超跑的提速能力与越野的持久耐 力，在不同任务场景实现动态效率最优。 复杂万卡集群的长稳运行 ： 为突破计算 - 内存动态博弈（服务区分配）、异构任务资源争抢 （车祸变道）及硬件耐久性（车道维护）瓶颈，需实现毫秒级资源再平衡与故障容错，保障作 业万级小时无间断运行。 基于仿真能力，通 过 全量部署空间搜索、动态性能感知调度优化等技术释放算力潜能 ，实现并行 配 ...