AI闹 . 和我们一起发现AI时代最重要的创造者 本文来自微信公众号： AI闹 ，作者：AI闹，原文标题：《AI投资中场战事：人才战、信仰分歧和超级机遇》，题图来自：AI生成 以下文章来源于AI闹 ，作者AI闹 半个月内，京东成了具身智能近期最大买家。 7月到8月，刘强东拍板，京东在极短时间窗口内连续扫货，把头部公司见了个遍。从5月底出手智元机器人开始，到这一波连续投千寻智 能、逐际动力、众擎机器人、RoboScience、帕西尼。 其中千寻、逐际河众擎三家，甚至是在同一天宣布完成交易。战火密集到让人来不及反应。 表面看来，京东的故事符合一个标准答案：智能机器人是下一波智能汽车级别的巨大浪潮，而京东必须抢占这个入口。宇树 Go2 具身机器 狗在京东平台上一个月就能卖2000只，具身的消费端才刚刚起步。 但也有一种市场观点，认为以上逻辑更像是"内部汇报版本"。老刘要投具身，团队要有落地页，必须找一个合理的说法。 真正的原因不得而知。 或许和同行有关。毕竟美团在具身及相关产业链上的布局甚早，大大小小已近30家。 无论如何，京东正在通过激进投资具身智能，全力提升自身科技标签中的AI成色。 而这只是2025 年中国A ...

Core Insights - The article discusses the emergence of Vision-Language-Action (VLA) models based on large Vision-Language Models (VLMs) as a transformative paradigm in robotic manipulation, addressing the limitations of traditional methods in unstructured environments [1][4][5] - It highlights the need for a structured classification framework to mitigate research fragmentation in the rapidly evolving VLA field [2] Group 1: New Paradigm in Robotic Manipulation - Robotic manipulation is a core challenge at the intersection of robotics and embodied AI, requiring deep understanding of visual and semantic cues in complex environments [4] - Traditional methods rely on predefined control strategies, which struggle in unstructured real-world scenarios, revealing limitations in scalability and generalization [4][5] - The advent of large VLMs has provided a revolutionary approach, enabling robots to interpret high-level human instructions and generalize to unseen objects and scenes [5][10] Group 2: VLA Model Definition and Classification - VLA models are defined as systems that utilize a large VLM to understand visual observations and natural language instructions, followed by a reasoning process that generates robotic actions [6][7] - VLA models are categorized into two main types: Monolithic Models and Hierarchical Models, each with distinct architectures and functionalities [7][8] Group 3: Monolithic Models - Monolithic VLA models can be implemented in single-system or dual-system architectures, integrating perception and action generation into a unified framework [14][15] - Single-system models process all modalities together, while dual-system models separate reflective reasoning from reactive behavior, enhancing efficiency [15][16] Group 4: Hierarchical Models - Hierarchical models consist of a planner and a policy, allowing for independent operation and modular design, which enhances flexibility in task execution [43] - These models can be further divided into Planner-Only and Planner+Policy categories, with the former focusing solely on planning and the latter integrating action execution [43][44] Group 5: Advancements in VLA Models - Recent advancements in VLA models include enhancements in perception modalities, such as 3D and 4D perception, as well as the integration of tactile and auditory information [22][23][24] - Efforts to improve reasoning capabilities and generalization abilities are crucial for enabling VLA models to perform complex tasks in diverse environments [25][26] Group 6: Performance Optimization - Performance optimization in VLA models focuses on enhancing inference efficiency through architectural adjustments, parameter optimization, and inference acceleration techniques [28][29][30] - Dual-system models have emerged to balance deep reasoning with real-time action generation, facilitating smoother deployment in real-world scenarios [35] Group 7: Future Directions - Future research directions include the integration of memory mechanisms, 4D perception, efficient adaptation, and multi-agent collaboration to further enhance VLA model capabilities [1][6]

Core Insights - The "2025 Inclusion·Bund Conference" will take place from September 10 to 13 in Shanghai, focusing on the theme of "Reshaping Innovative Growth" and gathering global tech leaders, young scientists, entrepreneurs, and tech enthusiasts [1][5] Event Highlights - The conference will feature a series of activities including one main forum, over 40 open insight forums, 18 innovation stages, and three AI technology competitions, along with a new venture capital Meetup and a 5,000 square meter "Robot Town" [5][11] - The event has attracted nearly 20,000 young tech talents globally, with participation in various activities such as the innovation stage and AI competitions [7][11] Notable Participants - The conference will host 16 academicians and Turing Award winners, including Richard Sutton and Jack Dongarra, along with over 550 industry experts and business leaders [7] - Young AI tech talents will play a significant role, with nearly 100 AI scientists and tech entrepreneurs participating [7] AI Competitions - The AI competitions will focus on three main areas: AI smart hardware, financial intelligence, and AI security, with over 8,000 teams from nearly 20 countries registered [8] Technology Exhibition - The conference will feature a 10,000 square meter cutting-edge technology exhibition showcasing advanced technologies such as large models and controllable nuclear fusion [10] - A "Robot Town" will be established, featuring over 40 renowned domestic and international embodied intelligence companies and more than 100 types of intelligent robots [10] Recruitment and Networking - The event will also include a technology talent recruitment fair with over 1,500 job openings from more than 50 companies, alongside an industry matchmaking event attracting over 100 companies [11]

Core Insights - Tesla's "Master Plan Part IV" emphasizes the integration of AI into physical products and services through robotics and automation, aiming for a sustainable and prosperous future [1][2] - Elon Musk stated that approximately 80% of Tesla's value will derive from the Optimus robot, highlighting its significance alongside the expansion of Tesla's full self-driving software [1][2] Summary by Sections Introduction - Tesla's guiding principle since its inception has been to achieve unrestricted sustainability without compromise, with a new mission to bring AI into the physical world [1] - The plan outlines five guiding principles: infinite growth, innovation eliminating limitations, using technology to solve real problems, automation benefiting humanity, and broader accessibility leading to greater growth [1][2] Key Principles - **Infinite Growth**: Growth in one area does not necessitate decline in another; resource scarcity can be addressed through technological advancements [1] - **Innovation Eliminating Limitations**: Tesla has historically challenged norms, such as the transition from fossil fuel vehicles to electric ones, proving the feasibility of low-cost battery production [1] - **Using Technology to Solve Real Problems**: Each Tesla product aims to be more efficient and sustainable than its predecessor, with the Optimus robot redefining labor and enabling humans to focus on preferred activities [1] - **Automation Benefiting Humanity**: The development of autonomous driving should enhance human living conditions, with Musk envisioning a transition from a "pre-autonomous" to a "post-autonomous" world [1] - **Broader Accessibility Leading to Greater Growth**: To create a prosperous society, Tesla aims to produce advanced products that are affordable and scalable, expanding its business from clean energy to intelligent systems that enhance labor and mobility [2] Conclusion - The plan concludes with a vision of tools that will help build a sustainable and prosperous world, redefining labor, mobility, and energy to benefit all [3] Reactions - Media and public reactions to the plan have been mixed, with some praising its ambitious vision while others criticize it as utopian and lacking clarity compared to previous plans [5][6]

Core Viewpoint - The company emphasizes the importance of R&D in PEEK modified materials and has developed key polymer material solutions for lightweight bionic skeletons, self-lubricating joint bearings, and high-temperature motor housings [1] Group 1: R&D and Material Development - The company has organized a dedicated R&D team to tackle challenges in PEEK modified materials, resulting in the development of PEEK-CF50, PEEK-SL20, and PEEK-HT300 [1] - Innovations include the addition of 50% carbon fiber, PTFE + graphite, and ceramic fiber + carbon fiber to enhance material performance, achieving characteristics such as lightweight, self-lubrication, and high-temperature resistance [1] Group 2: Business Strategy and Collaborations - By the first half of 2025, the company plans to leverage its advantages in automotive injection mold R&D, design, and manufacturing, and actively engage in technical cooperation with domestic humanoid robot companies [1] - Currently, the company is developing supporting products for a leading domestic humanoid robot company, with the R&D project progressing steadily as planned [1]

Group 1 - The core focus of Tesla is shifting from its automotive business to the development of humanoid robots, with CEO Elon Musk stating that approximately 80% of the company's future value will come from the Optimus robot [1] - Tesla's Master Plan Part IV emphasizes the integration of artificial intelligence into products and services, aiming to create a sustainable and prosperous future [1] - The release of Master Plan Part IV marks a strategic shift for Tesla, moving away from electric vehicles and energy towards artificial intelligence and robotics [1] Group 2 - Despite the strategic shift, Tesla's electric vehicle sales have been declining in multiple markets, with a reported 4% year-on-year decrease in wholesale sales in China for August [2] - In California, Tesla's electric vehicle registrations fell by 21.1% in Q2, marking the seventh consecutive quarter of decline, which is critical as California accounts for over 30% of the U.S. zero-emission vehicle registrations [3] - In Europe, Tesla's sales have also been weak, with a reported 84% year-on-year drop in vehicle registrations in Sweden for August and a 47.3% decline in France during the same period [3]

Core Viewpoint - The article discusses the current state of the A-share market, highlighting the potential for a bull market driven by technology sectors, while also noting the presence of bubbles in certain segments that may require adjustments [5][8][10]. Market Overview - On September 2, the three major indices in A-shares collectively declined, with the Shanghai Composite Index falling by 0.45%, the Shenzhen Component down by 2.14%, and the ChiNext Index dropping by 2.85% [13]. - The total trading volume for the day reached 29,127.66 billion yuan, an increase of 1,347.62 billion yuan compared to the previous day, with over 1,200 stocks rising [13]. Sector Performance - The technology sector, particularly in AI-related fields, is highlighted as a key driver of the current market rally, with significant capital expenditures from major U.S. companies boosting domestic firms' performance [9][10]. - The solid-state battery sector is experiencing heightened activity, with multiple upcoming industry conferences and government support expected to further stimulate market performance [13][14]. Investment Trends - Morgan Stanley reports that while the market shows signs of overheating, regulatory intervention is not yet anticipated. The trading volume on August 26 reached a record high of 3.2 trillion yuan, indicating increased market activity [10]. - The article notes a significant inflow of retail investor participation, with daily net inflows averaging 11 billion yuan, although this remains below previous averages [10]. Notable Events - A rumor regarding Tesla's optimistic production guidance for robots has led to a surge in related stocks, although the authenticity of the information remains unverified [12][14]. - The article mentions upcoming events in the solid-state battery sector, including several key conferences scheduled for September and October, which are expected to influence market dynamics [14]. Capital Flow - The article indicates that the banking, public utilities, and household appliances sectors are leading in capital inflows, while the communication, computer, and electronics sectors are experiencing outflows [26][38]. - Northbound capital transactions totaled 3,670.85 billion yuan, with significant trading in stocks like ZTE Corporation and China Merchants Bank [29][30].

9月2日，宇树科技宣布，预计将在2025年10月至12月期间向证券交易所提交上市申请文件，届时公司的相关运营数据将正式披露。若成功上市，将成为A 股"人形机器人第一股"，验证赛道商业化可行性。 此前，在2025世界机器人大会上，宇树科技创始人兼首席执行官王兴兴表示："我们公司已成立九年，正在推动上市流程。我把公司上市比作高考，这是 企业迈向更成熟管理和运营的重要阶段。"他进一步说明："上市不仅是对过去九年发展的总结，更是对股东的负责。这将推动公司进入新的发展阶段。 宇树科技在公告中披露，以2024年为例，四足机器人、人形机器人和组件产品的销售额分别占约65%、30%和5%。其中，约80%的四足机器人被应用于 研究、教育和消费领域，而剩余的20%则被用于工业领域，如检查与消防。人形机器人完全用于研究、教育和消费领域。 值得注意的是，虽然当前营收贡献仍以四足机器人为主，但宇树的人形机器人正逐渐成为产业关注的焦点。2025年亮相央视春晚不仅一夜走红，更是点燃 了产业发展的引擎。7月11日，宇树中标人形机器人"第一大单"，中国移动采购与招标网公示的结果显示，智元机器人和宇树科技分别中标中移（杭州） 信息技术有限公司人 ...

Core Insights - Yushu Technology is actively preparing for its IPO, with plans to submit application documents between October and December 2025 [1] - The company has completed the counseling filing with the Zhejiang Securities Regulatory Bureau, and a comprehensive assessment will be conducted to determine if it meets the listing conditions [1][4] - The controlling shareholder, Wang Xingxing, holds 23.82% of the company's shares directly and controls an additional 10.94% through a partnership, totaling 34.76% [4] Financial Overview - In 2024, Yushu Technology's revenue structure is expected to consist of approximately 65% from quadruped robots, 30% from humanoid robots, and 5% from component products [4] - About 80% of quadruped robots are utilized in research, education, and consumer sectors, while the remaining 20% are applied in industrial fields such as inspection and firefighting [5] Strategic Vision - Wang Xingxing views the IPO as a significant step for the company, likening it to a high school examination that signifies maturity in management and operations [4] - The company emphasizes its commitment to applying high-performance general-purpose robots across various civilian industries since its establishment [5]

机器人板块午后拉升，截至收盘，国证机器人产业指数上涨1.9%，中证智能电动汽车指数下跌0.6%， 中证消费电子主题指数下跌3.2%，中证物联网主题指数下跌3.5%，机器人ETF易方达（159530）今日获 2.03亿份净申购。 国金证券指出，2025H1机器人板块主业业绩向好，120家公司共实现营收2888.25亿元，同比增长 13.81%，其中近九成公司实现盈利。随着智元、宇树、优必选等人形机器人企业斩获大额订单，机器 人商业化进展正在加速，产业链协同不断深化，行业有望进入发展快车道。 | 机器人ETF易方达 | | | 159530 | | --- | --- | --- | --- | | 跟踪国证机器人产业指数 | | | | | | | | .. /////////// | | | ® H | 该指数 | 该指数自2019年 | | 该指数聚焦机器人产业,包 | 该指数涨跌 | 发动市销半 | 以来估值分位 | | 括机器人本体、核心零部件 | | | | | 等环节,其中人形机器人" 含量"高、占比近80%。 | 1.9% | 5.2倍 | 99.7% | | 智能汽车ETF易方达 = | | | ...