2026年上海市政府工作报告将"加快建设免陪护照护服务体系""深化多元复合医保支付方式改革，加强创新药和医 疗器械推广应用"等作为2026年的主要任务。 "此次我的发言建议是'免陪照护服务'背后的专业人才岗位和体系打造。"2026年上海两会召开期间，2月3日下 午，在上海市政协十四届四次会议一场医卫界别分组会议（下称"分组会议"）上，上海市政协常委、复旦大学附 属中山医院（下称"中山医院"）副院长钱菊英表示，随着老龄化进程加速，"免陪照护服务"能在未来进一步减轻 独生子女负担，"目前，我院平均每天会为1300至1400位病人开设此项服务，但具备对应技能的护理人员较为缺 乏、受教育程度普遍不高、劳动关系较为分散，这些问题亟需引起重视"。 "'药'的创新、转化与应用，不但关乎民众生命健康，也是上海'十五五'规划建议中促进生物医药高端高效发展的 核心任务所在。"上海市政协委员、上海交大系统生物医学教育部重点实验室主任韩泽广在分组会议上如是说，生 物医药领域的创新，则需要"产学研用"全链条足够顺畅。 上述委员们对于卫生健康、生物医药领域的建言，不但能为大众关心的民生事项提供"纾解"，也正为政策制定"前 置"提供进一步 ...

Core Insights - The interview with Li Yuanqing, a former Huawei talent, focuses on the potential for China to create its first large-scale embodied intelligence product and the importance of "multi-machine heterogeneity" as a future direction [1] Group 1: Market Trends and Developments - By 2025, the embodied intelligence sector is expected to see significant growth, driven by major tech companies and startups securing funding, indicating a long-term market logic [3] - The linkage between primary and secondary markets is evident, with listed companies investing in robotics to enhance traditional manufacturing and create new growth avenues [3] - The maturity of technology in the sector is improving, which is crucial for sustaining high market interest [3] Group 2: Technological Advancements - The performance of humanoid robots has significantly improved, with capabilities to withstand physical interactions and perform complex tasks, showcasing advancements in technology [5] - The development of large models has led to a qualitative change in the intelligence of embodied systems, with success rates for simple tasks increasing from 60% to 100% [5] Group 3: Data Challenges and Solutions - A major bottleneck in the industry is the scarcity of high-quality, large-scale physical interaction data, which is costly to collect [8] - Simulation-generated data and data factories are emerging as key solutions, with a "data pyramid" framework explaining their roles in data generation and application [8][10] - The core value of world models lies in efficiently generating foundational data to support model training, addressing the need for diverse data [10] Group 4: Cost and Implementation Challenges - The high costs of essential components, such as industrial computers and robotic hands, pose significant barriers to the widespread adoption of embodied intelligence [13] - The lack of clarity in defining application scenarios for humanoid robots further complicates the assessment of their return on investment [13] Group 5: Future Directions and Opportunities - Li Yuanqing advocates for a multi-machine heterogeneity approach, where different types of robots collaborate to complete complex tasks, reflecting a natural ecosystem of specialization [15] - The competitive edge for Chinese companies in 2026 will hinge on product deployment and data integration, with the potential for the first widely adopted embodied intelligence product to emerge from China [15] - The current environment presents a favorable opportunity for entrepreneurs and researchers to engage in this sector, with expectations for rapid technological maturation and cost reduction [15]

Core Insights - The industry is shifting focus from the design of robotic bodies and algorithm iterations to infrastructure development for data production [1][4] - The establishment of the Super EID Factory by Shenzhen Pasini aims to provide large-scale, high-quality multimodal training data, addressing the shortage of tactile data essential for enhancing robotic skills and generalization capabilities [1][3] Group 1: Infrastructure Development - The Super EID Factory covers nearly 12,000 square meters and is expected to produce nearly 200 million high-quality training data entries annually [1] - The factory employs a "no-body dependency" data collection system with 150 standardized units to capture human hand movements, spatial trajectories, and mechanical interaction information in real scenarios [1][2] - This approach is designed to significantly reduce data production costs and enhance the versatility of the data produced, making it applicable across various robotic configurations [1][2] Group 2: Technological Innovations - The factory utilizes proprietary "Neural Mesh" technology to synchronize and fuse high-precision tactile data with visual, joint angle, and voice information, creating rich high-dimensional data streams [2] - The "Soma Redirect" system allows the collected human data to be effectively adapted to different robot structures, addressing the long-standing challenge of model generalization across different robotic bodies [2] Group 3: Industry Trends - Various infrastructure development paths are emerging in the embodied intelligence sector, including Pasini's third-party data service factory model and Shanghai's Zhiyuan Robotics' vertical integration strategy [2][3] - In Beijing, a collaborative effort among government and leading enterprises is focused on building an "AI public computing power platform" and industry datasets to support local businesses [3] - The Guangdong province's innovation center aims to integrate resources from universities and the industry to establish a shared data collection and management mechanism [3] Group 4: Strategic Goals - Companies are aiming beyond merely being "data suppliers"; for instance, Pasini plans to use the factory's data to build an "OmniSharing DB" and create a growth flywheel with its self-developed large models [3] - The ultimate goal is to construct a "world model" that deeply understands the laws of the physical world and to open the factory's data capabilities to the global industrial ecosystem [3][4] Group 5: Competitive Landscape - The emergence of embodied intelligence data factories signals a transition from theory to practice and from prototypes to products, indicating a deepening of industry competition [4] - Competition is evolving to encompass not just algorithms or hardware but also data production, model training, and vertical integration capabilities [4]