【导读】阿里年报披露变革成绩单，阿里致股东信聚焦AI 中国基金报记者 曹雯璟 6月26日晚间，阿里巴巴集团发布2025财年年报，再次交出一份积极的变革成绩单：在AI需求的强劲推动下，阿里云财年收入突破双位数 增长，AI相关产品收入连续七个季度实现三位数同比增长。淘天集团投入用户增长和体验提升，新买家强劲增长，88VIP会员规模超5000 万。跨境业务表现强劲，国际数字商业集团财年收入同比增长29%。 在致股东信中，阿里巴巴表示，阿里的基因里没有守成，只有创造。今天的阿里巴巴，正在以创业者的姿态，开启面向AI时代的全新征程。 净利润同比增长77% 阿里年报披露变革成绩单 年报显示，2025财年（2024年4月1日至2025年3月31日），阿里巴巴集团收入9963.47亿元，净利润同比增长77%至1259.76亿元。 年报显示，2025财年，阿里云服务了约63%的中国A股上市公司，阿里云模型平台"百炼"的用户数量呈现显著增长，阿里AI产品正在传统 垂直行业迅速渗透。 财年内，淘天集团客户管理收入（CMR）同比加速增长至6%。高质量消费群体88VIP会员规模增长至超5000万人，并继续展现出高留存 率。年报显示，2 ...

" 未来 5 到 10 年内，我们有 50% 的概率实现通用人工智能（AGI）。 " Google DeepMind CEO、 诺贝尔化学奖得主 Demis Hassabis 在本月初的一个采访中给出了惊人的预测。 AI 浪潮正以前所未有的速度加速进行中，在这条被追捧、被加速、甚至被神化 的 AGI 之路上，我们是否也踩过无数"看似正确却最终失败"的岔路？ 早在 2019 年，"强化学习之父"Richard Sutton 发表了著名的《The Bitter Lesson》（苦涩的教训）一文，他指出：过去 70 年，AI 研究最大的弯路 就是我们太依赖人类已有的知识与直觉，而历史一再表明，真正能带来突破的，反而是那些"摒弃人类经验、依靠计算规模"的方法。 在 Sutton 看来，算力与数据才是推动 AI 向前的根本引擎，而不是人类引以为傲的"智慧"本身。 这句话听起来也许刺耳，却道出了一个时代的真相——在这场智能革命中，许多我们曾深信不疑的"路径"与"经验"，正在变成新纪元的障碍。 这也正是我们本期《万有引力》直播对话的主题——「 通往 AGI 之路的苦涩教训」 。 6 月 27 日 12:00-14:00，C ...

联合创始人兼材料总监朱德航毕业于美国宾夕法尼亚大学，专注于仿生皮肤硅胶材料技术研发，曾荣获中国化 学奥林匹克竞赛一等奖，其在材料科学领域的专业积累为机器人仿生外观的突破提供关键支撑。 机器人大讲堂获悉，超仿生角色人形机器人初创企业「首形科技」近日官宣成功完成 Pre A轮融资。 本轮融 资由招商局创投、深创投共同领投，五源资本、浔商创投跟投，其具体融资金额尚未对外披露。值得注意的 是， 首形科技此前在刚刚成立仅一个月的时候便获得了德迅投资、奇绩创坛、 Taihill、智元机器人的投资支 持。 据机器人大讲堂了解， 首形科技【 全称：首形科技（上海）有限公司 】 成立于 2024年6月， 是一家专注 于高端超仿生人形机器人研发的企业。公司 致力于将尖端人工智能技术与类人机器人技术深度融合至服务机 器人领域，打造兼具高度交互能力与逼真外观的人形机器人产品。 作为类人机器人领域的探索者，首形科技以 "引领世界走向更智能、更和谐的未来"为愿景，持续深耕人工智 能与机器人技术的核心领域，涵盖电机驱动、大型语言模型、机器人运动控制、表示学习、自我监督学习及革 命性仿生皮肤技术等方向。通过技术创新，公司旨在推动机器人智能化 ...

Core Viewpoint - The partnership between OpenAI and Microsoft is facing unprecedented challenges as they negotiate the definition of "Artificial General Intelligence" (AGI), which involves significant commercial interests and control over AI technology [1][3]. Group 1: Investment and Agreements - Microsoft invested $1 billion in OpenAI in 2019, followed by an additional $13 billion in subsequent agreements, which included terms that allow OpenAI to restrict Microsoft's access to future technologies once AGI is achieved [3]. - The agreements aim to ensure OpenAI's technological independence and prevent monopolization by a single commercial entity [3]. Group 2: Definition and Standards of AGI - There is currently no clear technical definition of AGI, but it is generally believed to require cross-domain generality, autonomous evolution, consciousness, and self-awareness [3]. - OpenAI favors a "functionalism" definition based on the system's actual capabilities, while Microsoft advocates for a "gradual assessment" approach to avoid premature triggering of technology restriction clauses [3][4]. Group 3: Implications for Microsoft - Microsoft's reliance on OpenAI's technology is deeply integrated into its core business, with Azure providing computational support and OpenAI's API embedded in tools like Microsoft Copilot and GitHub Copilot [4]. - If OpenAI restricts technology sharing upon achieving AGI, Microsoft could face technological obsolescence, strategic disadvantages, and financial losses [4]. Group 4: Negotiation Stalemate - Negotiations between OpenAI and Microsoft have reached an impasse, with Microsoft proposing to transfer AGI determination rights to a third-party organization, which OpenAI has rejected due to concerns over compromising technical confidentiality [4]. - OpenAI is forming an independent committee with academic and ethical experts to establish AGI determination standards and plans to release a technical white paper by 2026 [4].

近年来，人工智能领域每隔一段时间就会出现一个突破性进展。技术层面，通用人工智能 （AGI）、量子计算与脑机接口等领域已无现成经验可循，人类正在踏入一个个科技领域的"无人 区"。产业层面，人工智能在不同场景下与不同技术的结合，正在为人类的未来带来无穷想象。而 技术狂飙下，人类与人工智能的关系应当如何调整？《两说》以"探索AI'无人区'"为题，推出一组 系列对谈，邀请资深嘉宾，就人工智能发展过程中的热点问题进行深入探讨。 本期节目，由第一财经主持人张媛对话2013年诺贝尔化学奖获得者、斯坦福大学结构生物学系 终身教授、复旦大学复杂体系多尺度研究院荣誉院长迈克尔·莱维特。在这位计算生物学家眼中， 地球上存在的三种智能是如何相互影响的？生物智能对人类有怎样的启示？机器智能（人工智能） 带来哪些颠覆性变革？面对生物智能和机器智能，人类智能的独特价值是什么？莱维特教授将从跨 学科的视角分享他的观点和洞察。 0 1 生物智能的深刻启示 迈克尔·莱维特是计算生物学领域的专家，研究横跨生物学、物理学、化学、计算科学等多个学科，为 当今世界的科研和创新提供了独特视角，也因此获得诺贝尔化学奖。在这位计算生物学家眼中，地球上存 在三 ...

Group 1 - The gaming sector is experiencing a notable rebound, with the gaming ETF (159869) showing a near 1% increase and potentially achieving a four-day winning streak [1] - In May, China's gaming market grew by 9.86% year-on-year, reaching a market size of 280.51 billion yuan, driven by new and mature products [2] - Mobile games saw a year-on-year growth of 11.96%, while self-developed games' overseas revenue increased by 8.3%, indicating robust growth dynamics [2] Group 2 - The release of GPT-5 by OpenAI is expected to enhance industry-leading model performance and improve application sentiment, marking a significant step towards achieving Artificial General Intelligence (AGI) [1] - The gaming sector is catalyzed by multiple factors including AI, content, and commercialization model transformations, suggesting potential investment opportunities in the gaming ETF (159869) [2]

Core Viewpoint - The relationship between OpenAI and Microsoft is deteriorating over the definition of Artificial General Intelligence (AGI) and related contract terms, leading to significant disputes that could impact OpenAI's upcoming IPO [1][4]. Group 1: Contract Disputes - OpenAI and Microsoft signed a $10 billion cooperation agreement in 2023, which has triggered major disagreements over the AGI clause [1]. - Microsoft is demanding the removal of a key clause that allows OpenAI to terminate technology access upon achieving AGI, but negotiations remain stalled as of May 2025 [1][2]. - The original 2019 agreement allows OpenAI to cut off Microsoft's access to technology if AGI is reached, a concept that some Microsoft executives view as unrealistic [1][2]. Group 2: AGI Development and Perspectives - OpenAI's CEO Altman claims that AGI is "within reach" and has defined it as a system capable of solving complex human-level problems across multiple domains [2]. - In contrast, Microsoft CEO Nadella argues that a 10% annual growth in the global economy is a more realistic benchmark, suggesting that AGI will not be achieved before the contract expires in 2030 [2]. Group 3: Financial Implications and Negotiation Dynamics - OpenAI is projected to burn through $46 billion in R&D over the next four years, making IPO financing a critical necessity [4]. - If the restructuring fails, the unique equity structure of OpenAI could jeopardize its IPO prospects [4]. - Microsoft has rejected several concessions from OpenAI, including giving up a 20% revenue share and allowing customers to access OpenAI models through competing cloud providers [3]. Group 4: Alliance Fractures - The partnership has seen a shift in power dynamics since the unexpected success of ChatGPT in late 2022, leading to visible fractures in their collaboration [7]. - OpenAI has turned to Google Cloud for computing resources due to insufficient support from Microsoft and has also partnered with Oracle [7]. - Microsoft is accelerating the development of its own AI models and recruiting teams to create alternative solutions [7].

Core Viewpoint - The company, Out of the Door, has launched its new AI hardware product, TicNote, which features advanced AI capabilities and aims to serve as a software carrier for continuous upgrades and iterations [2][7]. Product Launch - Out of the Door released TicNote, the world's first Agentic AI hardware, which is only 3mm thick and can be magnetically attached to smartphones [2]. - TicNote supports over 120 languages for transcription and can record continuously for over 20 hours [2]. - The overseas version was launched at a price of $159.99, while the domestic versions are priced at 999 yuan and 1499 yuan [2]. Company Strategy - The CEO, Li Zhifei, emphasized that the company is not focused on the sales volume of TicNote but rather on using it as a platform for software development and iteration [3][8]. - The company aims to avoid direct competition with major smartphone manufacturers by focusing on niche products that do not attract significant competition [9]. Financial Performance - For the fiscal year 2024, the company reported a revenue of 390 million yuan, a decrease of 23% from the previous year [7]. - The operating loss was 207 million yuan, with a gross margin of 52.1%, down from 64.3% the previous year [7]. - Revenue from AI enterprise solutions plummeted by 92.7% to 16.42 million yuan [7]. Future Outlook - Li Zhifei indicated that the company plans to release additional products, including smartwatches and earbuds, that will also utilize the Shadow AI technology [2]. - The company is currently seeking new narratives to revitalize its market presence, focusing on familiar AI hardware and software applications [7][10].

贾皓钧将AI for Science（AI赋能科学研究）分为三个阶段：第一阶段是数据降维与分析，通过机器学习 处理高维科研数据，揭示变量间的隐藏关联；第二阶段是科学性假设生成与实验验证的智能化。比如顶 尖科学家可能花一天提出假设，其他人用十年验证，AI则可以同时优化这两部分；第三阶段则接近通 用人工智能（AGI），AI或能独立发现如牛顿定律级别的科学规律。贾皓钧认为，这会比AGI晚一步， 但会很快到来。 贾皓钧还分享道，对于垂直领域的模型来说，高质量的数据可能比模型本身更为重要。"模型的性能其 实是两方面决定的，第一是这个模型本身架构好还是不好，第二是数据到模型训练好还是不好。垂直领 域尤其是科学领域，大家对结果精度要求非常高，不容忍90%的正确，必须100%严谨。" 专题：2025第十六届夏季达沃斯论坛 2025年新领军者年会于6月24-26日在天津举办，本届新领军者年会主题为"新时代企业家精神"。会议期 间，微博达沃斯特别主持于超对话深度原理（Deep Principle）创始人兼CEO贾皓钧。 对话中，贾皓钧回忆称，公司最初源于2022年麻省理工学院（MIT）的一个康复类创业项目，后转向AI 赋能化学与 ...

以下文章来源于东四十条资本 ，作者王满华 东四十条资本 . 聚焦股权投资行业人物、事件、数据、研究、政策解读，提供专业视角和深度洞见 | 创投圈有趣的灵魂 将投中网设为"星标⭐"，第一时间收获最新推送 估值将飙升至290亿美元。 作者丨 王满华 来源丨 东四十条资本 这恐怕又将成为Y Combinator投资历史上的经典一役。 据The Information消息，Facebook的母公司Meta将以148亿美元（约合人民币1062.9亿元）收 购Scale AI 49% 的股权，这笔交易将成为Meta继190亿美元收购WhatsApp之后的第二大交易。 Scale AI是一家数据标注公司，谷歌、微软、OpenAI、xAI等Meta的直接竞争对手，都是Scale AI 的大客户。该公司去年新一轮融资后的估值约为138亿美元，而Meta入股后，其估值飙升至290亿 美元，直接实现翻倍。 近期，美国科技并购市场持续升温，一些明星AI项目都获得了来自买方的高溢价收购要约，这为项目 背后的早期投资人贡献了不少回报百倍的退出案例。 今年28岁的Alexandr Wang成长于美国新墨西哥州洛斯阿拉莫斯的一个华裔家庭，父 ...