【环球时报驻德国特约记者 昭东 环球时报记者 李迅典 环球时报特约记者 陈欣】编者的话：奥地利《标准报》26日报道称，鉴于欧洲与美国的争 端愈演愈烈，严重依赖美国科技服务的欧洲，正加速推动"数字主权"，以防美科技巨头的各项服务被美国政府下令关闭。然而，缺乏相关科技基 础且市场被美国巨头牢牢占据的欧洲想要实现"数字脱钩"并不容易。一些欧美媒体和接受《环球时报》记者采访的专家认为，开展多元化合作才 是欧洲在全球数字生态中真正实现战略自主的有效路径。 "欧洲不能继续充当'数字租户'" "欧洲正为噩梦般的局面做准备，即美国封锁数字服务。"美国《华尔街日报》报道称，特朗普政府暂时撤回了此前"必要时将武力接管格陵兰 岛"的威胁。即便如此，这也为欧洲注入了紧迫感——如何减少对支撑其经济诸多领域的美国数字基础设施和工具的依赖。对欧洲官员而言，最坏 的情况是什么？是白宫发布行政命令，切断欧洲企业与政府正常运转所必需的数据中心或电子邮件的访问权限。 法国24小时新闻台报道称，欧盟与美国之间日益紧张的关系，暴露了欧洲最脆弱的领域之一：对美国大型科技公司的高度依赖。 奥地利《标准报》称，德国信息学会公开警告，鉴于美国的"2025年国 ...

去年1月20日，DeepSeek正式发布R1模型震惊全球。转眼一年过去，中国的"AI崛起"已成为有目共睹的 现实。 国际舆论场上，关于AI竞争中美谁更胜一筹的话题沸腾不息。近日，美国总统特朗普语出惊人，称在 其带领下，"美国在AI领域领先，并远远领先中国。" 现实究竟如何，我们又该怎样看待这场竞赛？ （一） 美国这波AI热潮形成突破的"技术奇点"，本质上是"大力出奇迹"，即大幅扩大模型的参数规模、训练数 据规模等，从而让这个模型"智能涌现"。 在规模化法则（Scaling law）被默认为大模型的第一性原理之后，AI竞赛迅速变成了算力竞赛。因此我 们看到，美国科技界掀起了AI芯片、数据中心投资热潮，芯片巨头英伟达的股价就此一飞冲天；美国 政府也将芯片"武器化"，通过各种出口禁令来遏制他国AI产业的发展。 然而，通过不停堆算力来提升AI的智能水平是否真的是唯一发展路径？ 目前，已有大量公开论文指出，规模化法则正在遭遇严重的"边际效应递减"问题，没有证据表明随着计 算资源的超常持续投入，复杂巨系统就一定能在某个临界点再次触发能力"非线性突变"，撞出下一 个"智能涌现"的时刻。 因此，除了豪赌算力资源，进行成本高昂 ...

分组1 - Tech health stocks are nearing record highs with significant earnings expected from major companies [1][10] - Amazon is closing its physical stores while increasing focus on delivery services, impacting competitors like Walmart, Uber, and DoorDash [16][18] - A $6 billion partnership between Meta and Corning aims to enhance fiber-optic cable support for Meta's data centers, resulting in a 17% stock increase for Meta [2][4] 分组2 - Micron is investing $24 billion over ten years in Singapore to address memory supply shortages, particularly for AI applications [4][5] - SK Hynix is experiencing stock gains due to anticipated benefits from a U.S. free trade agreement and Microsoft's AI chip initiatives [3][6] - The memory market is tightening, affecting smartphone and PC manufacturers as they adapt to increased demand for flash memory [5][12] 分组3 - Northwood has raised $100 million in funding to enhance satellite capabilities, indicating strong demand for its phased array antennas [45][48] - The company has secured a $49 million contract with the U.S. Space Force, demonstrating its operational capabilities [49][50] - Northwood plans to expand its global presence with 18 new sites across five continents [56] 分组4 - Anthropic has released a new version of its Claude chatbot, which can perform actions on computers, potentially impacting white-collar jobs [34][36] - The AI landscape is shifting towards specialized intelligence, with significant investments expected in this area [60][62] - Companies like Harvey are focusing on specialized AI applications, indicating a trend towards niche expertise in the AI sector [62][68]

A group of about 50 AI experts laid off by Sony Israel has moved to the Israel offices of British chip giant ARM, "Globes" has learned. Sony Israel, based in Hod Hasharon, announced in November that it was shutting down its R&D office in Israel after 10 years. To support its employees, Sony promised to help finance the re-establishment of Altair, the Israeli startup it acquired in 2016, which was then developing 4G cellular communication chips. It was believed that only some of the 200 employees would be ...

Group 1 - The core message from Tencent's annual meeting is a pragmatic acknowledgment of past delays and a focus on AI as the sole growth engine worth significant investment [2] - Tencent's leadership is shifting from a cost-cutting approach to a more aggressive stance, indicating a need to compete directly with ByteDance [2] - The current state of Tencent's AI is fragmented, with a disconnect between AI model development and product application, leading to internal chaos [3][4] Group 2 - The experience of OpenAI and Google suggests that the AI era requires integration of research and product development, rather than maintaining separate research institutions [5] - Tencent risks falling behind if it cannot produce competitive native AI applications, similar to the challenges faced by Google against OpenAI [5] - Tencent and Alibaba have struggled to develop truly consumer-oriented products, unlike ByteDance, which has effectively utilized technology and rapid iteration to dominate the market [6]

根据这项试验，参与的诊所将获得人工智能系统，该系统旨在辅助患者分诊、症状分析和医疗记录。这 些工具有望帮助临床医生更高效地整理患者信息，并标记出需要密切关注的潜在风险。卢卫生官员表 示，这些系统将严格作为决策支持工具运行，最终的诊断和治疗决定仍由训练有素的专业人员做出。支 持者认为，即使节省的时间不多，也能显著改善医疗机构的护理服务，因为在这些机构中，1名护士1天 可能要照看几十名患者。 卢旺达试点项目是"地平线1000"计划的一部分，该计划是由盖茨基金会与OpenAI合作支持的一项为期 两年、总额5000万美元的倡议。该项目旨在到2028年为非洲各地多达1000家基层医疗诊所提供支持。盖 茨在宣布这项计划时表示，人工智能可能成为面临卫生工作者短缺和国际援助减少的国家的重要工具。 他认为，如果负责任地部署人工智能，可以帮助保护在孕产妇保健、疫苗接种覆盖率和传染病监测等领 域来之不易的公共卫生成果。 尽管人们对该计划持乐观态度，但专家警告说，成功实施将取决于本地化和信任。目前大多数人工智能 系统主要以英语运行，而卢旺达语是患者和许多一线工作人员的主要语言。数字政策专家警告说，语言 支持不足可能会限制其在实际应用 ...

Core Viewpoint - Tesla is expected to report weak Q4 2025 earnings, with revenue projected to decline by 3.6% to $24.5 billion and profit per share potentially dropping by 40% to $0.44, driven by a 15.6% decrease in vehicle deliveries. However, the energy segment may provide a significant upside surprise, potentially offsetting some of the automotive challenges [1]. Group 1: Financial Performance Expectations - Revenue is anticipated to fall to $24.5 billion, a 3.6% decrease year-over-year [1]. - Profit per share is expected to decline by 40% to $0.44, reflecting a significant drop in vehicle deliveries [1]. - Vehicle deliveries are projected to decrease by 15.6%, indicating a slowdown in Tesla's auto business [1]. Group 2: Energy Segment Insights - Tesla's Energy segment is expected to generate $3.825 billion in revenue with a gross margin of approximately 31.1%, nearly double the automotive gross margin of 17% [1]. - The energy revenue grew by 44% year-over-year in Q3, driven by demand from AI data centers requiring stable power [1]. - Tesla's Megapack is positioned as a solution for power supply volatility, catering to major clients like xAI, Google, and OpenAI [1]. Group 3: Challenges and Risks - Regulatory credit revenue fell by 43.6% year-over-year to $417 million, with forecasts predicting a 75% decline in 2026, impacting overall profitability [1]. - The political actions of CEO Elon Musk have been estimated to have cost Tesla between 1.0 to 1.26 million potential vehicle sales from October 2022 to April 2025, with significant impacts noted in Q1 2025 [1]. - The shift to a subscription model for Full Self-Driving (FSD) may lead to a surge in January sales but could face challenges in subscriber retention and monetization in subsequent quarters [1].

Core Insights - The article discusses the limitations of current AI systems, particularly their inability to execute tasks in the physical world despite strong logical reasoning capabilities, a phenomenon termed "physical illusion" [4][11] - A new framework called "WorldMind" is introduced, which allows AI to learn from execution errors and successful experiences, enhancing their ability to interact with the physical environment [5][10] Group 1: Physical Illusion in AI - AI systems often generate plans that are logically sound but physically unexecutable, highlighting a disconnect between semantic reasoning and physical world understanding [4][11] - This issue is likened to a novice cook who can describe cooking steps but fails to perform them correctly due to a lack of practical experience [4] Group 2: WorldMind Framework - WorldMind is based on the predictive coding theory from cognitive science, emphasizing active prediction and error correction as a learning process [5] - The framework consists of two main components: process experience, which learns from execution errors, and goal experience, which extracts effective strategies from successful task completions [5][6] Group 3: Testing and Results - The effectiveness of WorldMind was validated in two testing environments, EB-ALFRED and EB-Habitat, showing significant improvements in task success rates, with EB-ALFRED increasing from 44.4% to 48.0% and EB-Habitat from 43.6% to 48.8% [7] - WorldMind demonstrated cross-model transfer capabilities, allowing experiences learned by one AI model to assist different AI models, thus promoting shared learning [7][10] Group 4: Broader Implications - The research suggests a shift towards AI systems that continuously learn and accumulate experiences rather than relying solely on static knowledge [9][10] - This approach aligns with human cognitive characteristics and may lead to more robust AI systems capable of adapting to dynamic environments [10]

英伟达以每股87.20美元的价格购买了价值20亿美元的CoreWeave A类普通股，这是继9月份63亿美元投 资之后的又一笔大手笔投资。 Q&A 近几个月来，CoreWeave一直在积极扩张业务。9月份，该公司与Meta签署了142亿美元的协议，并扩大 了与OpenAI的合作合同，据报道价值224亿美元。 最新的投资还将帮助测试和验证CoreWeave的AI原生软件和参考架构，包括SUNK和CoreWeave任务控 制系统，并使这些产品能够纳入英伟达为云和企业客户提供的参考架构中。 CoreWeave联合创始人、董事长兼首席执行官迈克尔·因特拉托表示："当软件、基础设施和运营协同设 计时，AI才能成功。这种扩大的合作突显了我们在客户群中看到的强劲需求，以及AI系统转向大规模 生产时更广泛的市场信号。" 9月份，CoreWeave参与了一项价值150亿美元的英国AI基础设施建设计划，该计划还包括Nscale、英伟 达和OpenAI。该计划包括在全球部署30万块英伟达GPU。 然而，HyperFrame Research分析师兼首席执行官史蒂文·迪肯斯对CoreWeave向最终用户交付价值的能力 仍存疑问。迪肯 ...

以下文章来源于海豚研究 ，作者海豚君 海豚研究 . 有灵魂的思考，有态度的研究！ 本文来自微信公众号： 海豚研究 ，作者： 海豚君，原文标题：《深扒Minimax与智谱：大模型，一 场算力强度与融资耐力的残酷绝杀？》，头图来 自：视觉中国 2026年初ChatGPT发布三年之际，中国两大AI大模型创业公司Minimax和智谱几乎同时，均以约60 亿美金的估值上市，且上市后都以暴涨带动整个AI应用的普涨行情。但同时，扒开两个大模型公司 的报表，都是"亏无止境"的面相。 一边是全球大模型疯狂卷价格、出现越来越多的大宗商品 ， 一边是模型无底洞的研发投入，但同时 还有这种上市即翻倍的盛景。冰火交织下，海豚君一直的疑问是大模型到底是一种什么样的生意。 这次，海豚君就结合两个上市的公司数据，来认真探讨一下这个问题： 大模型到底需要怎样的投入要素和密度？ 以下是详细分析： 一、拔高的收入、惊悚的投入 两个大模型公司Minimax和智谱，都是"短小精悍"型——人手少、产品迭代快，收入增长快。人数到 2025年下半年没超一千人，收入零起步，两三年年化收入都在快速迈向一亿美金。 但到了支出层，再高增的收入，在凶猛的投入面前也相 ...