相关事件 五一视界(06651.HK)：SimOne4.0已成功完成在摩尔线程MTTS5000GPU上的系统性适配与深度优化 港 股异动丨AI应用股集体走强，五一视界、MINIMAX-WP涨超5% 摩尔线程MTTS5000是专为大模型训练、推理及高性能计算而设计的全功能GPU智算卡，基于第四代 MUSA架构"平湖"打造。其单卡AI稠密算力最高可达1000TFLOPS，完整支持从FP8到FP64的全精度计 算。 公司SimOne4.0和摩尔线程MTTS5000GPU的成功适配，为智能驾驶行业生态(包括主机厂、一级供应 商、检测机构等)提供了完整的国产化解决方案，证明了国产GPU进入了自动驾驶高精度、高负载任务 的阶段。与此同时，2026年新版《道路机动车辆生产企业准入审查要求》对机动车生产准入审查升级， 要求智能驾驶相关车企必须具备仿真、封闭场地、实际道路三类验证能力，并拥有安全性评估能力。高 阶智能驾驶车进入量产的同时，仿真测试正在成为智能驾驶车上路前的刚需。 格隆汇2月23日丨五一视界(06651.HK)公告，近日，51Sim下一代智能驾驶与机器人仿真平台SimOne4.0 已成功完成在摩尔线程MTTS50 ...

香港交易及結算所有限公司及香港聯合交易所有限公司對本公告的內容概不負責，對其準確性 或完整性亦不發表任何聲明，並明確表示概不就因本公告全部或任何部分內容而產生或因倚賴 該等內容而引致的任何損失承擔任何責任。 Beijing 51WORLD Digital Twin Technology Co., Ltd. 北京五一視界數字孿生科技股份有限公司 （於中華人民共和國註冊成立的股份有限公司） 自願性公告－業務發展最新情況 （股份代號：06651） 北京五一視界數字孿生科技股份有限公司（「本公司」）董事會（「董事會」）欣然宣 布，近日，51Sim下一代智能駕駛與機器人仿真平台SimOne4.0已成功完成在摩爾 線程MTT S5000 GPU上的系統性適配與深度優化，從大模型感知挖掘、4DGS模 型訓練到4DGS仿真推理和合成數據生成，全面支撐端到端、VLA和世界模型等 多條智駕技術路線的量產落地，進而擴展至機器人等具身智能領域的訓練和智能 升級，實現了物理AI高置信度閉環仿真與合成數據的全棧國產化。 51Sim作為本公司的三大業務之一，為中國領先的合成數據與仿真平台。其核心 產品包括SimOne（智駕仿真平台）、Da ...

Core Insights - The article highlights the privacy risks associated with AI plugins, revealing that over half of the sampled Chrome AI plugins collect user data, with nearly one-third targeting personally identifiable information (PII) [1][3]. Group 1: AI Plugin Data Collection - A study by Incogni analyzed 442 AI-labeled plugins, finding that many use scripting permissions to access user input and alter webpage content [3]. - High-risk categories include programming assistants, math tools, meeting assistants, and voice transcription plugins, with notable examples being Grammarly and Quillbot [3]. - The current trend in AI deployment relies heavily on cloud services, making AI plugins a convenient way for users to access AI capabilities without complex installations [3]. Group 2: Data Scarcity and AI Development - The article discusses a looming "data drought" for AI companies, with predictions that high-quality text data on the internet will be exhausted by 2028, and machine learning datasets may run out by 2026 [5]. - The reliance on synthetic data has emerged as a solution, but it has proven inadequate in practical applications, leading to issues like underfitting and model failures [5]. - Media and content platforms are becoming aware of the value of their data, leading to legal battles with AI companies over data usage rights [5]. Group 3: Privacy Concerns and User Choices - The article raises concerns about the lack of regulation for browser plugins compared to stricter app stores, allowing malicious plugins to bypass oversight [7]. - AI plugins are primarily distributed through personal blogs, AI community links, and GitHub, as developers prioritize efficiency over regulatory compliance [9]. - Users face a dilemma of whether to trade privacy for convenience, with over 50% of AI plugins collecting user data, making it a widespread issue [12].

Core Insights - The report from Beijing Zhiyuan Artificial Intelligence Research Institute outlines the top ten AI technology trends for 2026, highlighting advancements in multimodal AI, embodied intelligence, and multi-agent systems [1][3][4]. Group 1: Multimodal AI and World Models - In 2025, discussions around multimodal AI surged, with expectations for 2026 to see further exploration of world models that can simulate real-world laws, enhancing AI's understanding of physical concepts [3][4]. - The value of world models lies in their ability to mimic human cognitive processes, enabling AI to tackle problems that are simple for humans but challenging for machines [3]. Group 2: Embodied Intelligence - As of 2025, over 230 companies in China are focused on embodied intelligence, with more than 100 in humanoid robotics, indicating a significant industry presence [4]. - The report anticipates a potential reshuffling in the embodied intelligence sector due to global economic uncertainties, with companies needing to adapt to evolving foundational models [4]. - Humanoid robots are expected to advance into real-world applications, with examples like Tesla Robotics' Optimus 2.5 being utilized in various operational settings [4]. Group 3: Multi-Agent Systems - The transition from single-agent to multi-agent systems is seen as essential for adapting to complex workflows, with multi-agent systems demonstrating advantages in handling intricate tasks [5]. - Communication protocols among agents are expected to mature, facilitating practical applications in production environments by 2026 [5]. Group 4: AI in Scientific Research - The emergence of AI Scientists capable of executing complete research processes marks a significant shift in scientific discovery, driven by foundational models and automated experimental facilities [6]. - The U.S. has initiated the "Genesis Mission" to enhance AI's role in scientific research through integrated platforms and efficient data sharing mechanisms [6]. Group 5: AI for Science in China - China faces challenges in the AI for Science domain, particularly in computational power, data, and model infrastructure, despite its relative advantage in AI applications [7]. - Progress is being made with the establishment of a national scientific data sharing platform, but there is a need for improved scientific foundational models [7]. Group 6: Personal and Industry Applications - The rapid development of AI personal applications in 2025 has led to the rise of "AI super applications," which integrate multiple services for users [8]. - Industry applications are still in exploratory phases, with more complex AI agents facing challenges such as data quality and system integration [8]. Group 7: Synthetic Data and AI Safety - The shift towards synthetic data is anticipated as high-quality data resources dwindle, with the synthetic data market in China growing significantly from 1.18 billion to 4.76 billion in four years [10]. - AI safety concerns are rising, with reports indicating that leading models struggle with preventing misuse, prompting the industry to develop new security frameworks [11].

Group 1 - The core focus of the article is on the future of artificial intelligence (AI) and the potential reshaping of the Chinese internet landscape, particularly in the context of embodied intelligence and the emergence of new super apps [1][19][20] - The embodied intelligence sector is expected to undergo a significant reshuffle by 2026, with over 230 companies currently in the market, including more than 100 humanoid robot firms [2][4] - The industry is transitioning from experimental phases to mass production, with humanoid robot sales surpassing 10,000 units, indicating a move towards initial commercialization [4][9] Group 2 - Major players in the global humanoid robot market include Zhiyuan, Yushun, and Ubtech, with Zhiyuan leading with an annual shipment of over 5,100 units, capturing 39% of the global market share [6] - The competition for AI super apps is intensifying, with companies like OpenAI and Alibaba striving to create comprehensive platforms that integrate various services into a single interface [10][11] - The article highlights the emergence of vertical players in specific domains, such as Ant Group's "Antifor Health" and MiniMax's AI companionship applications, which have successfully established business models and attracted significant user bases [14] Group 3 - The AI industry is predicted to enter a "trough of disillusionment" in 2026, with many AI projects failing to deliver measurable impacts, leading to a potential delay in AI investments [15][19] - The market for synthetic data is experiencing rapid growth, with projections indicating that by 2030, the scale of synthetic data will surpass that of real data, becoming the primary fuel for model training [17] - The advancements in data quality and the evolution of AI technologies are expected to drive significant changes in the operational dynamics of the real economy [19]

Core Insights - The article discusses the transition of artificial intelligence (AI) from digital applications to physical applications, marking a significant shift in 2026 towards "physical AI" [1][4][12] Group 1: Development of Physical AI - Physical AI is characterized by its ability to understand physical laws and interact with the real environment, enabling a new research paradigm of "hypothesis - AI simulation - experimental verification" [4] - The development of physical AI is expected to take 5 to 10 years of deep cultivation, indicating a long-term investment in this area [4] - The concept of "world models" is crucial for physical AI, allowing intelligent agents to simulate actions in a virtual environment before executing them in reality, which is essential for safety and efficiency [5][6] Group 2: Data Generation and Training - The industry is leveraging "synthetic data" generated from high-fidelity physical simulation engines to train AI models at zero marginal cost, although there remains a challenge in bridging the "simulation to reality" gap [7] - A promising approach involves using human daily videos for pre-training AI models, allowing them to learn physical common sense and operational skills from real-world scenarios [7] - The future of physical AI data solutions may involve a "trinity" ecosystem composed of human experience, virtual simulation, and physical interaction [7] Group 3: Global Competitive Landscape - The development of physical AI shows a contrast between the U.S. and China, with the U.S. leading in foundational algorithms and cutting-edge exploration, while China excels in engineering and rapid deployment of technologies [9][10] - China's strategy emphasizes cost-effectiveness and clear application scenarios, supported by government initiatives that integrate AI into various sectors, setting ambitious goals for technology adoption [10] Group 4: Challenges and Future Directions - The ultimate goal of physical AI is to achieve generalization, enabling intelligent agents to adapt quickly to new environments and tasks, which remains a significant challenge [11] - Issues such as explainability, safety redundancy, and ethical standards are becoming increasingly important in the physical AI era, as the consequences of errors can have real-world implications [11] - The year 2026 is seen as a milestone, marking the beginning of AI's transition from virtual to physical applications, with ongoing advancements expected [12]

Core Insights - AI has transitioned from a laboratory concept to an omnipresent tool that can write articles, compose music, design, program, schedule in enterprises, inspect in factories, teach in classrooms, and diagnose in hospitals, effectively reducing costs for businesses and creating opportunities for entrepreneurs and investors [1][5] Industry Trends - The logic of profitability has shifted from "scaling" to "efficiency," with AI becoming the commercial core that addresses pain points across various industries, supported by a new synergy among policy, capital, industry, and social acceptance [3][7] - The current phase of AI integration into industries is the third stage, where service applications are central to AI's value release [3][7] - Future technological integrations, such as blockchain with AI, quantum computing with AI, and brain-computer interfaces with AI, are expected to create new business opportunities and potentially lead to the emergence of unicorn companies [3][7] AI Demand and Data Trends - Global AI computing power demand is projected to reach 10^23 FLOPS by 2024, which is 1 million times the total global computing power in 2010, and is expected to grow to 10^26 FLOPS by 2027, a 1000-fold increase in three years [3][7] - Data is viewed as the "oil" of AI, with four key trends anticipated: 1. Data assetization will become a core strategy for companies, with over 50% of listed companies expected to disclose data asset values in their financial reports by 2026 2. The data factor market will mature, transitioning from non-standard to standardized trading 3. Privacy computing technologies like federated learning and differential privacy will be widely adopted to address the "data usable but invisible" issue 4. Synthetic data will become a significant supplement, with its share in AI training expected to exceed 25% by 2027 [3][7] AI Implementation Framework - A five-layer architecture for AI implementation has been proposed, encompassing resource access, data management, Data & AI engineering, intelligent applications, and security operations, which has shown significant effectiveness in large and medium-sized enterprises, reducing project delivery cycles by over 50% and greatly increasing customer renewal rates [4][8] - It is anticipated that over 80% of large and medium-sized enterprises will adopt similar architectural frameworks to build AI infrastructure in the next three years [4][8]

Core Insights - The global attitude towards AI has shifted from "apocalyptic fears" to focusing on "releasing real industrial potential" by 2025, indicating a significant change in governance priorities [1][3] Group 1: Macro Landscape - The Paris AI Action Summit in February 2025 marked a shift from "safety anxiety" to "innovation and action," reflecting a restructuring of global governance logic [2] - The EU is adjusting its regulatory approach by introducing the "Digital Omnibus" proposal to simplify rules and delay high-risk obligations to enhance industrial competitiveness [2] - The U.S. is moving towards deregulation, with the Trump administration's focus on a unified federal framework to eliminate barriers for the industry [2] - China emphasizes a pragmatic approach, balancing specific regulatory measures with an application-oriented strategy, creating a layered governance system [2] Group 2: Data Governance - The AI industry faces a structural shortage of high-quality data, leading to a search for synthetic data as a key solution [4] - Legislative efforts in the EU and Japan are establishing frameworks for "text and data mining," while U.S. court rulings are leaning towards recognizing the use of legally acquired books for training as "fair use" [4] - Future regulations may evolve beyond simple prohibitions to create a commercially viable mechanism for balancing rights and technological advancement [4] Group 3: Model Governance - The U.S. is shifting from comprehensive coverage to targeted regulation, exemplified by California's SB 53 law focusing on transparency for only a few large-scale models [7] - The EU's complex regulatory framework is facing challenges due to high compliance costs, prompting frequent legislative adjustments [7] - China's "scene slicing" strategy involves penetrating regulation across specific AI services, creating a governance system from data to application [7] - The rise of open-source models like DeepSeek-R1 is reshaping the global AI landscape, highlighting the importance of establishing a "safe harbor" for contributors [8] Group 4: Application Scenarios - The transition of AI from cloud to real-world applications raises new privacy challenges, particularly with intelligent agents that require extensive permissions [10] - AI's evolution into emotional companions introduces risks related to emotional dependency, prompting diverse regulatory approaches to protect vulnerable groups [10] - The struggle against deepfakes highlights the limitations of watermarking technologies, suggesting a focus on high-risk scenarios for precise governance [11] Group 5: Future Outlook - The discussion around AI consciousness and welfare is evolving from philosophical debates to scientific validation, indicating a potential need for governance frameworks that address AI as a rights-bearing entity [13]

Core Viewpoint - The core viewpoint of the article is that a significant paradigm shift is occurring in artificial intelligence (AI), moving from a focus on language learning and parameter scale to a deeper understanding and modeling of the physical world, as highlighted in the 2026 AI technology trends report by the Beijing Zhiyuan Artificial Intelligence Research Institute [2][5]. Summary by Sections AI Technology Trends - The competition in foundational models is shifting from the size of parameters to the ability to understand how the world operates, marking a transition from "predicting the next word" to "predicting the next state of the world" [5][9]. - The year 2026 is identified as a critical turning point for AI, transitioning from the digital world to the physical world, driven by three main lines: cognitive paradigm elevation, embodiment and socialization of intelligence, and dual-track application value realization [8]. Key Trends - **Trend 1: World Models and Next-State Prediction** There is a consensus in the industry moving towards multi-modal world models that understand physical laws, with the NSP paradigm indicating AI's mastery of temporal continuity and causal relationships [9]. - **Trend 2: Embodied Intelligence** Embodied intelligence is moving from laboratory demonstrations to real industrial applications, with humanoid robots expected to transition to actual production and service scenarios by 2026 [10]. - **Trend 3: Multi-Agent Systems** The resolution of complex problems relies on multi-agent collaboration, with the standardization of communication protocols like MCP and A2A enabling agents to work together effectively [11]. - **Trend 4: AI Scientists** AI is evolving from a supportive tool to an autonomous researcher, significantly accelerating the development of new materials and drugs through the integration of scientific foundational models and automated laboratories [12]. - **Trend 5: New "BAT" in AI** The C-end AI super application is becoming a focal point for tech giants, with companies like OpenAI and Google leading the way in creating integrated intelligent assistants, while domestic players like ByteDance and Alibaba are also actively building their ecosystems [13]. - **Trend 6: Enterprise AI Applications** After a phase of concept validation, enterprise AI applications are entering a "disillusionment valley," but improvements in data governance and toolchains are expected to lead to measurable MVP products in vertical industries by the second half of 2026 [15]. - **Trend 7: Rise of Synthetic Data** As high-quality real data becomes scarce, synthetic data is emerging as a core resource for model training, particularly in fields like autonomous driving and robotics [16]. - **Trend 8: Optimization of Inference** Inference efficiency remains a key bottleneck for large-scale AI applications, with ongoing algorithmic innovations and hardware advancements driving down costs and improving energy efficiency [17]. - **Trend 9: Open Source Compiler Ecosystem** Building a compatible software stack for heterogeneous chips is crucial to breaking the monopoly on computing power, with platforms like Zhiyuan FlagOS aiming to create an open and inclusive AI computing foundation [18]. - **Trend 10: AI Safety** AI safety risks are evolving from "hallucinations" to more subtle "systemic deceptions," with various initiatives underway to enhance safety mechanisms and frameworks [19]. Conclusion - The Zhiyuan Research Institute emphasizes that the ten AI technology trends provide clear anchors for future technological exploration and industrial layout, aiming to promote a stable transition of AI towards value realization [21].

Core Insights - The article highlights the rapid growth of the embodied intelligence sector, with a significant demand for synthetic data, which is currently being met by the company Guanglun Intelligent, founded in 2023 [1][2][4]. Group 1: Company Overview - Guanglun Intelligent has positioned itself as a key player in the synthetic data market, providing AI simulation services that fill a critical gap in the industry [1][2]. - The company claims that over 80% of the simulation assets and synthetic data for leading international embodied intelligence teams come from them [1][7]. - The founder, Yang Haibo, emphasizes the importance of high-quality synthetic data for the development of embodied intelligence, which is expected to be as ubiquitous as smartphones and cars in various industries [2][18]. Group 2: Market Demand and Growth - The demand for synthetic data in the embodied intelligence sector is at least 1000 times greater than that for autonomous driving, driven by the need for complex physical interactions [2][8]. - Initially focused on the autonomous driving sector, the company has seen a surge in demand from the embodied intelligence and world model fields since mid-2024 [8][18]. - The industry has shifted from questioning the use of synthetic data to focusing on how to effectively produce it [6][8]. Group 3: Technical Challenges and Solutions - The main challenges in generating high-quality synthetic data include ensuring physical accuracy and adapting to evolving data requirements from embodied models [10][12]. - The company employs a proprietary "solve-measure-generate" approach to simulation, which allows for precise modeling of complex physical interactions [11][12]. - The training process for synthetic data must balance quality and scalability, with the company aiming to produce large volumes of high-quality data [10][14]. Group 4: Future Outlook - The company envisions itself as a foundational infrastructure provider for the physical AI era, focusing on continuous development of simulation technologies [18]. - The industry is expected to transition from a tool-based phase to a foundational industry phase, with a growing reliance on reliable data support for the widespread application of robots and intelligent agents [18].