现在的问题是，这些规模最终能存的住么，持仓的这些投资者能赚到钱么？这是需要回答的问题，还有就是监管对这些现象意识到问题了没 有，很多时候市场不理性的时候，我们要学会看到问题，这样才能更好的保护自己。 结论很简单，对于热门的赛道ETF投资的时候悠着点，一定要控制好比例，千万别抱着发财的梦想。 免责声明：文中内容仅供参考，不构成任何操作建议或提示，股市有风险，投资请谨慎！ 商业航天昨天出现了重挫，相关的ETF跌幅在5%左右，当然这个跌幅并不算大，有比这更夸张的，有些航天航空相关的ETF单日跌幅在9%， 而还有刚刚起来的AI应用方面，比如说个别的科创创业人工智能ETF单日下跌超过11%。 看到这样的波动，不知道大家作何感想？似乎也在警告我们，最近主题赛道似乎有点过热了，我看到有人对过往时间的主题赛道的涨幅和表 现做了一个统计，可能唯一一个表现时间长、涨幅明显的就是商业航天了，大多数主题赛道的表现都很一般，这样说其实已经很客气了，简 单说几乎都是从哪里起来的又回到了那里，表现的时间周期很短，等你刚刚明白过来的时候，结果短期就见顶了。 | 名称 | 现价 - | 涨跌幅 | 估算规模(亿元) | 跟踪指数名称 | | ...

Core Insights - Elon Musk predicts that Starlink will transport 300 to 500 GW of solar photovoltaic components to space annually for AI computing, potentially exceeding the total computing power of the United States within two years [1] - Musk emphasizes the rapid advancement of AI, forecasting that by 2030, AI intelligence will surpass the combined intelligence of all humans [3] - He asserts that within three years, robotic surgeons will outperform top human surgeons, rendering medical school obsolete [4][5] Group 1: AI Advancements - Musk describes the exponential growth of AI, stating that breakthroughs occur so rapidly that he is frequently astonished [3] - He predicts the arrival of AGI (Artificial General Intelligence) by 2026, with AI intelligence surpassing human capabilities by 2030 [3] - The potential for robotic doctors to provide superior medical care is attributed to three factors: AI capability growth, chip performance improvements, and mechanical dexterity advancements [4] Group 2: US-China AI Competition - Musk highlights that China is set to surpass other regions in AI computing power, citing three main advantages: significant electricity generation capacity, diminishing chip performance gaps, and unmatched execution speed of Chinese engineers [5][6][7] - By 2026, China's electricity generation is expected to reach three times that of the US, with a substantial portion derived from solar energy [6] - The decline of Moore's Law indicates that the performance gap in chip technology is narrowing, making it easier for China to catch up [6] Group 3: Economic Predictions - Musk suggests that in 10 to 20 years, traditional concepts of money may become irrelevant, as AI and robotics will drastically reduce production costs, leading to a new economic paradigm of "Universal High Income" [8] - He warns of a tumultuous transition period over the next 3 to 7 years, characterized by radical changes and societal upheaval [8] Group 4: Energy Solutions - Musk advocates for solar energy as the key to human energy independence, proposing a three-step plan: improving existing grid efficiency, launching solar AI satellites into space, and establishing satellite manufacturing facilities on the Moon [9][10][11] - He emphasizes the vast potential of solar energy, arguing that utilizing solar power from space is more efficient than terrestrial nuclear fusion [9] Group 5: Future of Currency - Musk concludes that the future of currency will fundamentally be energy, as it will drive AI and enable the production of goods [13]

2026开年，全球AI竞赛场上，再次出现零售巨头的身影。 1月11日，沃尔玛与谷歌宣布，计划将沃尔玛及山姆会员店的商品整合进谷歌的Gemini。与此同时，谷 歌在全国零售联合会（NRF）大会上，正式发布通用商业协议（UCP），用于为谷歌搜索和Gemini的AI 模式提供智能购物能力。美国用户无需离开AI聊天界面，即可在Gemini的对话框中浏览商品并完成购 买。 作者 | 肖思佳 编辑 | 乔芊 在此之前，上一个把AI对话变成购物场景的，是OpenAI。2025年9月底，ChatGPT推出的"即时结 账"（Instant Checkout），跑通了从对话到下单的完整购物闭环。 从2022年底ChatGPT发布算起，短短三年时间，科技行业已经历多轮高频的"你追我赶"。而在经历过一 轮的搜索入口博弈之后，AI竞赛的焦点，正在进一步延伸至电子商务领域。 据纽约邮报报道，2025年11月底的黑色星期五，在这美国一年中最繁忙的购物日里，消费者借助人工智 能完成搜索、比价、筛选与决策，推动在线消费额达到创纪录的118亿美元。Adobe Analytics数据显 示，这一天美国的在线消费额比去年同期增长了9.1%。AI正逐 ...

Core Viewpoint - The article emphasizes the increasing demand for VLA (Variable Latency Algorithms) talent, particularly in the autonomous driving sector, highlighting the challenges faced in data collection and model optimization [2][3][4]. Group 1: VLA Demand and Challenges - There is a significant demand for VLA algorithms, especially for autonomous driving, as reflected in the job market and academic publications [2]. - Many practitioners express frustration over the difficulties in tuning VLA models and the complexities involved in data collection [3][4]. - The reliance on real machine data for effective model training is underscored, with many companies advocating for a "real machine data" approach despite its challenges [5][8]. Group 2: Learning and Practical Application - The article discusses the difficulties beginners face in integrating data, VLA models, training optimization, and deployment, with some struggling for months without success [8]. - A new course has been developed to address these challenges, providing practical tutorials and hands-on experience with VLA methods [10][11]. - The course covers a comprehensive curriculum, including hardware, data collection, VLA algorithms, and real machine experiments, aimed at enhancing learning efficiency [13]. Group 3: Course Details and Target Audience - The course is designed for individuals seeking practical experience in the VLA field, including students and professionals transitioning from traditional fields [21]. - Participants will receive a SO-100 robotic arm as part of the course, facilitating hands-on learning [14]. - The course schedule is outlined, with classes starting on December 30, 2025, and continuing into early 2026 [22].

Core Viewpoint - The article emphasizes the transformative impact of AI on industries, highlighting the shift from discussing AI as a concept to its practical implementation in products and services, particularly observed during the CES 2026 event in Las Vegas [2][3]. Group 1: AI Implementation and Industry Trends - The CES event showcased the integration of AI into various products, with companies like Nvidia and Samsung demonstrating advancements that support the notion that "everything can be AI-enabled" [4]. - Chinese brands are evolving from merely exporting products to a more comprehensive "value export" that includes technology narratives, brand communication, and ecosystem building, reflecting a deeper competitive edge [4][11]. - Lenovo's presentation at CES highlighted its "hybrid AI" strategy, showcasing a personal super-intelligent agent and a complete enterprise solution, marking a significant leap from product output to co-building technology standards and ecosystems [8][9]. Group 2: Insights from Silicon Valley - The visit to Silicon Valley focused on understanding the foundational aspects of innovation, emphasizing that true innovation requires an ecosystem that supports exploration and embraces uncertainty [14][18]. - Discussions with industry leaders revealed that the main barriers to AI implementation are often organizational restructuring rather than technical challenges, underscoring the importance of addressing business pain points [17]. - The experience at tech giants like Google and Meta illustrated the significance of a collaborative culture and the integration of AI with hardware, providing a clearer vision of future possibilities [17][19]. Group 3: Upcoming Opportunities in Germany - The upcoming Hannover Industrial Fair in Germany will focus on "AI in manufacturing," providing insights into smart manufacturing, industrial automation, and AI technologies, which are crucial for Chinese companies aiming to enhance their global competitiveness [30][31]. - The journey to Germany aims to explore the deep-rooted manufacturing practices and standards that can serve as a strategic reference for Chinese enterprises looking to expand into European markets [28][29].

Core Viewpoint - The "Artificial Intelligence +" Action Plan of Jiangsu Province aims to accelerate AI technology innovation and integration across various industries, enhancing productivity and driving economic growth. Group 1: Action Plan Overview - The plan leverages Jiangsu's strengths in industry, data, scenarios, and talent to promote AI technology innovation and application across multiple sectors [2] - Key goals include achieving over 70% penetration of new intelligent terminals and applications by 2027, over 90% by 2030, and establishing a leading AI innovation hub by 2035 [3] Group 2: Research and Development Focus - The plan prioritizes AI-driven scientific research, establishing key laboratories and innovation platforms to foster breakthroughs in various fields [4] - It aims to cultivate high-level talent in AI and enhance the capabilities of research institutions [4] Group 3: Industrial Upgrading - The initiative emphasizes the integration of AI into traditional industries, particularly manufacturing, to drive transformation and efficiency [5][6] - It includes the development of a comprehensive framework for AI-enabled manufacturing and support for small and medium enterprises in digital transformation [6] Group 4: Education and Talent Development - The plan outlines initiatives to integrate AI into education, including the establishment of AI-related programs in universities and the promotion of AI literacy among students [8] - It aims to create a robust ecosystem for AI talent development, including the establishment of AI colleges and research centers [8] Group 5: Healthcare Integration - The plan highlights the integration of AI in healthcare, with significant data management initiatives and the establishment of platforms for data sharing and application [9] - It aims to enhance the healthcare industry through AI technologies, improving medical devices and health services [9] Group 6: Data Infrastructure - The plan emphasizes the importance of high-quality data for AI development, with initiatives to create valuable data sets and support the data annotation industry [9] - It aims to establish over 1,000 high-quality data sets by 2027 to support AI applications across various sectors [9]

Core Insights - The article discusses the rapid advancement of AI agents, particularly in cancer research and oncology, highlighting their capabilities beyond traditional AI systems [3][4][6] - AI agents can autonomously optimize drug design, propose treatment strategies, and handle complex multi-step problems that previous AI systems could not address [3][4][27] Group 1: AI Agents Overview - AI agents differ from traditional AI systems by possessing "action capabilities," allowing them to perceive their environment, plan multi-step tasks, and execute complex workflows with minimal human intervention [8][14] - The integration of large language models (LLMs) with external tools enables AI agents to actively gather information, analyze data, and take actions rather than merely responding to commands [14] Group 2: Applications in Cancer Research - AI agents can autonomously generate research hypotheses, design experimental protocols, execute data analysis, and write academic papers, marking a significant shift towards fully automated research processes [17][15] - Multi-agent collaborative systems are emerging, where different AI agents simulate human research teams by taking on specific expert roles, enhancing problem-solving comprehensiveness and decision-making transparency [18] Group 3: Clinical Oncology Applications - In clinical settings, AI agents can integrate various medical data sources, support treatment decisions, and automate clinical trial matching, significantly improving efficiency and patient outcomes [22][20] - AI agents are capable of simulating human expert reasoning in image analysis, allowing for more complex clinical problem-solving [23] Group 4: Future Outlook and Challenges - The article outlines a three-phase process of "agentification" in cancer research and oncology, predicting a transition from current AI interfaces to fully integrated systems with autonomous capabilities [28][29] - Challenges include the need for new evaluation metrics for AI agents' performance, integration hurdles from research prototypes to clinical tools, and ethical considerations regarding the autonomy of AI systems [27][29]

Core Insights - The article discusses the advancements and challenges in the field of embodied intelligence and humanoid robots, highlighting the need for scalable production and systematic operations to support industry growth [2][3]. Group 1: Company Developments - SenseTime's co-founder Wang Xiaogang emphasizes the importance of comprehensive capabilities for establishing a foothold in the humanoid robot sector, indicating that the company is not without its shortcomings [2]. - The launch of the ACE embodied research paradigm and the open-source commercial application of the "Awakening World Model 3.0" are significant milestones for the company, addressing core pain points in embodied intelligence [2][3]. - The company has built a full-link technology system that effectively addresses issues such as data scarcity and generalization difficulties in the industry [2]. Group 2: Industry Standards and Challenges - Wang Xiaogang, who is also involved in the standardization committee for humanoid robots, identifies three main challenges in establishing a standard system: lack of data sharing standards, unclear safety responsibilities, and the need for improved quality standards [3]. - The industry is still in its early stages, requiring collaborative efforts across the sector to develop effective standards [3]. Group 3: Technological Innovations - The ACE paradigm introduces a human-centric approach to data collection, significantly enhancing data quality and reducing costs compared to traditional methods [12][14]. - The new paradigm allows for the collection of millions of hours of data annually, which is crucial for the development of effective embodied intelligence systems [12][14]. - The "Awakening World Model 3.0" integrates multi-modal understanding and predictive capabilities, marking a significant evolution in the field [19][22]. Group 4: Strategic Collaborations - The company has formed strategic partnerships with leading firms in various sectors, including hardware and cloud services, to create a comprehensive ecosystem for embodied intelligence [27][29]. - Collaborations with companies like Galaxy General aim to leverage each other's strengths in technology and production to overcome key technical challenges [29][31]. Group 5: Market Focus and Future Outlook - The company plans to focus on commercial and industrial applications in the next 3-5 years, with an emphasis on high-standard environments like front warehouses and retail storage [32]. - The potential for large-scale deployment in commercial services is highlighted, while industrial applications face challenges due to data sensitivity and low willingness to share [32]. - The company aims to develop a unified platform to support the development of both software and hardware in the industry, similar to NVIDIA's CUDA ecosystem [23].

新华财经上海1月14日电 2026年开年，人工智能的热度在持续升温：1月初，"工赋上海"创新大会在上 海举行，整个会场被人工智能产业链上下游的从业者挤得满满当当；江苏常州宣布设立总规模50亿元的 人工智能专项基金，支持优质项目和团队落地；安徽合肥启动国家人工智能应用中试基地，聚焦基层医 疗等场景，为"健康中国2030"提供技术支撑……在"十五五"开局之年，长三角正加速抢滩人工智能产业 新高地。 "西有模速空间，东有模力社区。"上海通过两大创新载体，贯通模型研发与下游应用的产业链条，并配 套设立三大先导产业母基金，叠加"算力券""模型券""语料券"等政策工具，持续降低企业创新成本。 在政府层面，这一趋势同样清晰可见：新年伊始，长三角多地政府领导的新年调研，纷纷将人工智能产 业园区作为"第一站"；多地"新年第一会"也不约而同地把"人工智能"设为核心议题。 华东师范大学城市发展研究院院长曾刚对记者表示，多地政府将人工智能作为开年调研重点，反映出对 这一战略性新赛道的高度共识。 各扬所长 打造AI产业高地 "长三角正以算力统筹为基础、平台共建为支撑、数据流通为纽带、政策协同为引领，系统推进人工智 能一体化发展。"安徽 ...

Core Insights - The article highlights the emergence of domestic AI models, particularly focusing on Xiaomi's MiMo-V2-Flash, which has shown significant advancements in efficiency and performance [1]. Group 1: MiMo-V2-Flash Overview - MiMo-V2-Flash was officially released and open-sourced on December 17, featuring a total parameter count of 309 billion and active parameters of 150 billion, utilizing a mixture of experts (MoE) architecture [1]. - The model achieves a remarkable inference speed of 150 tokens per second, with costs reduced to $0.1 per million tokens for input and $0.3 for output [1]. Group 2: Cost Efficiency Innovations - The model employs a hybrid sliding window attention mechanism, which significantly reduces the storage of KV cache by nearly six times while maintaining long text capabilities, supporting a context window of approximately 256k [9]. - MiMo-V2-Flash integrates a three-layer MTP setup, enhancing encoding task speed by about 2.5 times and addressing GPU idle time issues in small batch on-policy reinforcement learning [10]. - The model utilizes a multi-teacher online policy distillation (MOPD) approach, requiring only 1/50th of the computational power of traditional methods to achieve peak performance, allowing for faster model iterations and self-evolution [10]. Group 3: Competitive Positioning - MiMo-V2-Flash has achieved high rankings in benchmark tests, placing among the top two in the AIME 2025 mathematics competition and GPQA-Diamond science knowledge test, with a programming capability score of 73.4% in the SWE-bench Verified test [6][11]. - The model's performance is noted to be competitive with leading models, closely approaching GPT-5-High in programming tasks [6].