【观网财经丨智能早报 8月6日】 OpenAI六年来首次开源推理模型 当地时间8月5日，OpenAI发布了两款开源模型gpt-oss-120b和gpt-oss-20b。这是自2019年开源GPT-2以 来，OpenAI首次发布开源模型。据介绍，gpt-oss-120b总参数量为1170亿，激活参数为51亿，能够在单 个H100 GPU上运行，仅需80 GB内存，专为生产环境、通用应用和高推理需求的用例设计，既可以部 署在数据中心，也能在高端台式机和笔记本电脑上运行。而gpt-oss-20b总参数量为210亿，激活参数为 36亿，专门针对更低延迟、本地化或专业化使用场景优化，仅需16GB内存就能运行，这意味着大多数 现代台式机和笔记本电脑都能驾驭。 Anthropic发布Claude Opus 4.1，编程、研究、数据分析更强大 当地时间8月5日，Anthropic发布新模型Claude Opus 4.1。Anthropic公布的数据显示，Opus 4.1在编程能 力方面实现了实质性突破，在编程评估基准SWE-Bench Verified上，该模型得分达到74.5%，这一成绩 相比Claude Sonnet 3 ...

克雷西 发自 凹非寺 量子位 | 公众号 QbitAI 只需一句话，就能生成可实时交互的3D世界。 刚刚，谷歌DeepMind发布了 新一代通用世界模型Genie 3 。 性能上，Genie 3相比上一代大幅升级，支持 720P画质，每秒24帧实时导航，以及分钟级的一致性保持 。 | Genie 2 | Genie 3 | | --- | --- | | 360p | 720p | | 3D Environments | General | | Limited keyboard / mouse actions | Navigation; Promptable world events | | 10-20 seconds | Multiple minutes | | Not real time | Real time | 前DeepMind科学家、AI 3D生成创业者Tejas Kulkarni受邀体验了Genie 3。 他使用Genie 3，生成了长达57秒的城市高空漫游场景 （下图为节选） ： Tejas评价，Genie通用性强，还能学习物理，而且拥有强大的记忆力。 看了Tejas的测试，Reddit网友直 ...

Market Overview - On August 5, U.S. stock indices opened high but closed lower, with the Dow Jones down 61.9 points at 44111.74, a decline of 0.14%; the Nasdaq fell 137.03 points to 20916.55, down 0.65%; and the S&P 500 dropped 30.75 points to 6299.19, a decrease of 0.49% [1] - Major tech stocks mostly declined, with Facebook down over 1%, Microsoft down over 1%, Nvidia down nearly 1%, Apple down 0.21%, Google down 0.19%, Tesla down 0.17%, while Amazon rose nearly 1% [3] Pharmaceutical Sector - Pharmaceutical stocks saw significant gains, with Pfizer closing up over 5% and UnitedHealth Group rising over 4% [4] Chinese Stocks - The Nasdaq Golden Dragon China Index fell 0.56%, with mixed performance among popular Chinese stocks. Tianjing Bio surged over 15%, Zhihu rose over 6%, and Canadian Solar increased over 4%. However, companies like Lingzhong Portal fell nearly 20%, Yum China dropped over 6%, and Atour fell over 4% [4] Bond Market - U.S. Treasury yields generally rose, with the benchmark 10-year Treasury yield closing at 4.218% and the more rate-sensitive 2-year Treasury yield at 3.739% [5] Economic Indicators - The ISM reported that the U.S. services PMI for July was 50.1, below the market expectation of 51.5 and the previous month's 50.8, indicating near stagnation in service sector growth. Employment metrics fell from 47.2 to 46.4, the lowest since the pandemic began [6]

Group 1 - Google DeepMind has launched Genie 3, the first real-time interactive world model, leveraging over a decade of experience in simulation environments [1] - Genie 3 allows users to generate interactive 3D environments at 24 frames per second and 720p resolution, enabling immersive experiences similar to video games [1] - AI technology is increasingly integrated into various stages of game development, enhancing efficiency and expanding content supply [1] Group 2 - The penetration of AI in the game industrial development chain is expected to continue, with potential efficiency gains across all manufacturers [1] - Companies actively involved in the AI + game content sector include Kaiying Network, Giant Network, Shengtian Network, Fuchun Co., and Dihun Network [1]

01 2025年的美国广告行业迎来了一个不容忽视的拐点。根据eMarketer的数据，上半年美国数字广告的年增长率，09年金融危机之后，第一次跌破10%。 在过去二十年里，数字广告几乎成了商业世界最可靠的增长引擎。无论宏观经济如何波动，只要用户在线时长还在增加，广告预算就会源源不断地涌入。 于是，数字广告的曲线像一条笔直的斜线，从社交兴起到移动互联网，再到短视频浪潮，它几乎从未让投资人或广告主失望。 哪怕在2009年金融危机的至暗时刻，数字广告也只是稍微踩了一下刹车，随即继续狂飙。 但今年，这条曲线第一次显得无力。在连续 16 年双位数扩张后，如今的个位数增长未来将持续好几年。 行业并没有遇到灾难性的黑天鹅，电商还在卖货，用户依然刷屏，平台仍旧牢牢攥着注意力，然而增速却突然慢了下来。 数字广告不是被别人抢走了蛋糕，而是蛋糕本身看上去有些涨不动了。 02 这场预算迁徙的起点，是搜索广告的衰减。 搜索广告曾经是数字营销最确定的选择，因果关系清晰：用户输入关键词，广告匹配需求，点击带来转化，ROI可预测。 但这种逻辑正在被AI打破：搜索广告的基础逻辑是"查询—点击—转化"，每一步都可计量、可优化，CPC定价模型也依 ...

风险提示及免责条款 市场有风险，投资需谨慎。本文不构成个人投资建议，也未考虑到个别用户特殊的投资目标、财务状况或需要。用户应考虑本文中的任何 意见、观点或结论是否符合其特定状况。据此投资，责任自负。 谷歌DeepMind宣布推出第三代通用的世界模型Genie 3 ，可以生成前所未有的多样化交互式环境，给出 文本提示，Genie 3可以生成动态世界，可以以每秒24帧的速度实时导航，并以720p的分辨率保持几分 钟的一致性。与前代模型（如 Genie 1/2）和视频生成模型（如 Veo 2，Veo 3对直觉物理学的深刻理 解）相比，Genie 3是第一个允许实时交互的世界模型，同时 Genie 相比，其一致性和真实感也得到了 提升。 ...

AI Education Market - Competition in the AI education space is intensifying [1] - Google's NotebookLM is now available to younger users [1] Product Development - NotebookLM targets younger users [1]

Summary of Key Points from the Conference Call Industry Overview - The conference call discusses the performance of major US technology companies in Q2 2025, highlighting the impact of AI on their business models and market dynamics [1][3][7]. Core Insights and Arguments - **Strong Performance of Major Tech Companies**: Major tech stocks like Meta, Google, and Amazon reported better-than-expected earnings, indicating robust economic activity in the US. Meta's advertising and Amazon's e-commerce businesses showed significant growth [1][3][9]. - **AI Demand Surge**: There is a persistent high demand for AI computing power, with cloud providers emphasizing that downstream AI demand exceeds supply. Microsoft Azure's revenue grew by 39%, Google Cloud by 31.7%, and Amazon AWS by 17.5% [1][3][12]. - **AI Commercialization Progress**: The commercialization of AI is advancing steadily, with rapid growth in API calls related to AI products. AI is enhancing advertising algorithms and content engagement, driving revenue growth for Google and Meta [1][3][17]. - **Capital Expenditure Trends**: Tech giants are increasing their capital expenditure to address the supply-demand imbalance in data centers. Microsoft plans a 58.4% increase in capital expenditure for FY 2025, while Google raised its guidance from $75 billion to $85 billion [1][19][20]. Company-Specific Insights - **Meta**: Experienced an 11.5% stock price increase due to the success of its AI-driven advertising business, with a 22% net profit surprise [3][9]. - **Microsoft**: Stock rose by 8.28%, driven by accelerated growth in cloud services aided by AI, with a 3.6% revenue beat [3][9]. - **Google**: Stock increased by 1.73%, with both advertising and cloud services showing strong growth, although net profit was 14% below expectations due to losses in non-core operations [3][9]. - **Amazon**: The only major company to see a stock price decline (7%), attributed to slower AWS growth (17.5%) and declining profit margins [4][9][14]. Market Dynamics and Trends - **Valuation Levels**: The Nasdaq index's forward 12-month PE ratio is approximately 27.6, slightly below the five-year average of 28.2. However, some tech companies like Microsoft and Meta are trading at higher valuations than their historical averages [5][6]. - **AI Computing Sector Impact**: The AI computing sector has been a significant driver of stock performance, benefiting related fields such as cybersecurity and IT operations [6][7]. - **Emerging Competitors**: New companies like Oracle, Coreweave, and Databricks are emerging in the AI cloud service space, focusing on GPU-intensive computing and industry-specific AI services, leading to a trend towards multi-cloud deployments [3][16]. Important but Overlooked Content - **Supply Chain Challenges**: Amazon faces supply constraints in chip production and power supply, which may limit its ability to expand capital expenditures and infrastructure investments [14][15]. - **Long-term AI Potential**: Despite current challenges, Amazon's AWS may regain competitive advantage in the long run due to its cost optimization capabilities, particularly in the scaling phase of generative AI [15][16]. - **China Market Insights**: The performance of US tech giants in AI offers lessons for the Chinese market, emphasizing the importance of addressing AI computing demand and focusing on cost-effective models [21][22][23]. This summary encapsulates the key points discussed in the conference call, providing insights into the performance of major tech companies, the impact of AI on their business models, and the broader market dynamics.

Summary of Key Points from the Conference Call Industry Overview - The conference call discusses the performance of major US technology companies in Q2 2025, highlighting the impact of AI on their business models and market dynamics [1][3][7]. Core Insights and Arguments 1. **Strong Performance of Major Tech Companies**: Major tech stocks like Meta, Google, and Amazon reported better-than-expected earnings, indicating robust economic activity in the US [1][3]. 2. **AI Demand Surge**: There is a persistent high demand for AI computing power, with cloud providers emphasizing a supply-demand imbalance in AI services [1][3][12]. 3. **Capital Expenditure Increases**: Tech giants are raising their capital expenditure forecasts for 2025 and 2026 to address the growing demand for data centers [1][8][19]. 4. **AI Commercialization Progress**: Companies are making steady progress in AI commercialization, with significant increases in API usage and AI programming demand [1][3][17]. 5. **Market Share Dynamics**: Amazon's AWS is lagging behind Microsoft Azure and Google Cloud in growth rates, raising concerns about its competitive position in the AI landscape [1][4][13][14]. 6. **Valuation Levels**: The Nasdaq index's forward 12-month PE ratio is approximately 27.6, slightly below its five-year average, indicating some companies are still overvalued [5]. 7. **Impact of Tariff Policies**: Tariff policies are expected to have a delayed impact on the economy, with predictions of a 0.5% decline in US GDP due to tariffs [11]. Additional Important Content 1. **Emerging Competitors**: New companies like Oracle, Coreweave, and Databricks are emerging in the AI cloud service market, focusing on GPU-intensive computing [3][16]. 2. **AI Marketing Efficiency**: AI is significantly enhancing advertising algorithms, leading to increased revenue for companies like Google and Meta [1][17]. 3. **Investment in AI Infrastructure**: Companies are investing heavily in AI infrastructure, with Microsoft planning a 58.4% increase in capital expenditure for FY2025 [19][20]. 4. **Future Market Trends**: The AI era is expected to bring profound changes to the cloud computing market, with a potential shift towards multi-cloud deployments [16]. 5. **China Market Insights**: The developments in the US tech sector provide insights for the Chinese market, particularly in AI demand and the importance of cost-effective models [21][22]. This summary encapsulates the key points discussed in the conference call, providing a comprehensive overview of the current state and future outlook of the technology industry, particularly in relation to AI advancements.

"我们并不是在这里挑选赢家或输家。我们的目标是为所有联邦政府员工提供尽可能多的工具，最大限 度提高他们的工作效率。不同的使用场景将需要不同的工具。"GSA副局长斯蒂芬·埃希基安(Stephen Ehikian)表示。 凤凰网科技讯 北京时间8月5日，据彭博社报道，美国政府的中央采购部门将把OpenAI、谷歌以及 Anthropic纳入政府批准的AI供应商名单，此举将为AI技术在联邦文职机构的广泛应用铺平道路。 就在数天前，美国总统特朗普签署三项行政命令，以重塑政府在AI领域的作用，其中包括一项要求联 邦机构仅采购"无意识形态偏见"的大语言模型。(作者/箫雨) 美国总务管理局(GSA)将于当地时间周二宣布这一举措，通过其"多重采购计划"(MAS，一个预先谈妥 合同条款的联邦采购平台)提供这些AI工具，加快AI工具在联邦政府中的普及。若无此项便利措施，各 机构通常需要耗费数月时间自行协商技术使用条款。 GSA官员表示，已对三家公司的AI模型进行了多项性能与安全措施评估，也就是OpenAI的ChatGPT、 谷歌的Gemini和Anthropic的Claude。GSA未立即披露具体合同条款，但它此前曾利用自身的采购 ...