科技日报北京1月22日电 （记者付毅飞）22日，中科宇航将微重力金属增材制造返回式科学实验载荷交 付于中国科学院力学研究所。此前，这台由该所自主研制的载荷，搭载于中科宇航力鸿一号遥一飞行 器，于1月12日成功开展我国首次太空金属增材制造（即"3D打印"）实验。 据悉，这是我国首次基于火箭平台实施太空金属增材制造返回式科学实验。实验团队在太空微重力环境 下利用3D打印技术成功制备出金属零部件，整体技术达到世界一流水平。 任务过程中，科研人员突破了微重力条件下金属增材制造的物料稳定输运与成形、全流程闭环调控、载 荷—火箭高可靠协同等一系列关键技术。实验结束后，载荷舱经伞降系统平稳着陆回收。科研人员成功 获取了太空微重力环境中金属增材制造的过程数据，包括熔池动态特征、物料输运、凝固行为等；同 时，还获取了太空增材制造金属件的成形精度与力学性能等参数，为我国太空金属增材制造技术的快速 迭代积累了宝贵实验资料。 此次实验成功，标志着我国太空金属增材制造正式从"地面研究"阶段迈入"太空工程验证"新阶段，将有 力推动我国太空制造技术发展，为未来太空基础设施建设提供关键支撑。 ...

在运行中，MARS系统团队各司其职，通过自然语言交互实现任务规划、逻辑推理与决策制定，实 现了从任务规划—实验设计—代码编程—实验执行—数据分析的全流程闭环自主探索。 MARS系统展现出多AI与多机器人间的高效协同，在极短时间内实现微胶囊等功能性材料的快速创 制与性能优化，将原本4个月的研发时间压缩至4小时。 目前，该研究成果相关核心专利已转让给武汉中科先进材料科技有限公司实施应用。双方共建 的"材料中试智能"创新联合体获批首批国家级先进功能材料制造业中试平台，并建设有全国一流的微胶 囊中试产线。团队将MARS系统与微胶囊中试优化结合，已快速完成灭火微胶囊等多种功能产品的工艺 开发和优化，多个产品已走上货架。 1月22日，记者从中国科学院深圳先进技术研究院获悉，该院材料人工智能研究中心研究员喻学锋 团队打造出一支"AI科学家团队"——"多AI—多机器人"协同智能体系统（以下简称"MARS系统"），并 将其用于微胶囊（封装微球）等多种新材料的创制。相关研究成果1月21日发表在国际期刊《物质》 上。 新材料研发是一项涉及多学科知识交叉的复杂系统工程，通常面临周期长、成本高、流程繁琐等挑 战。 据悉，受人类研发团队 ...

Core Insights - The successful delivery of the microgravity metal additive manufacturing return scientific payload by China Aerospace Science and Technology Corporation marks a significant milestone in China's space manufacturing capabilities [1] - This experiment represents China's first implementation of space metal additive manufacturing based on a rocket platform, transitioning from ground research to space engineering validation [1] Group 1: Experiment Details - The payload was developed by the Institute of Mechanics, Chinese Academy of Sciences, and was successfully launched on January 12, conducting China's first space metal additive manufacturing experiment [1] - The experiment utilized 3D printing technology to successfully fabricate metal components in a microgravity environment, achieving world-class technical standards [1] - Key technological breakthroughs were made in material stability transport and forming under microgravity conditions, closed-loop control throughout the process, and high-reliability collaboration between the payload and rocket [1] Group 2: Data and Outcomes - The experiment successfully collected process data in the microgravity environment, including characteristics of the molten pool, material transport, and solidification behavior [1] - Parameters such as the forming accuracy and mechanical properties of the metal components produced in space were also obtained, providing valuable experimental data for the rapid iteration of China's space metal additive manufacturing technology [1] - The successful completion of this experiment is expected to significantly advance China's space manufacturing technology and provide critical support for future space infrastructure development [1]

长期以来，人们习惯于将全球水短缺称为一场"危机"，总认为这只是暂时的阵痛，终将迎来复苏。然 而，联合国大学20日发布的一份最新报告打破了这一幻想。 这份名为《全球水资源破产：后危机时代水资源匮乏现状》的报告指出，人们熟悉的"水资源紧 张"和"水资源危机"等术语，已无法反映当今许多地区的严峻现实。相反，后危机时代的特点是自然水 资源资本遭遇不可逆转的损失，系统已无法恢复到历史基线水平。这不仅是"流量"的匮乏，更是"存 量"的枯竭。 正如联合国大学水、环境与健康研究所所长卡维·马达尼所言："对世界大部分地区来说，'常态'已成过 去。"这不再是短期波动的阵痛，而是一个必须诚实面对的系统性失败，倒逼我们从传统的"危机应 对"彻底转向"破产管理"。 水系统陷入"资不抵债" 这份报告指出，世界已经进入"全球水资源破产"时代。 如何理解"全球水资源破产"?报告将其定义为"资不抵债"与"不可逆性"的交织。在这种状态下，长期的 取水量与污染负荷已双双突破可再生补给的红线。 当水系统陷入破产，人类的"饭碗"首当其冲。报告数据显示，全球约70%的淡水抽取用于农业，而地下 水则支撑了全球约50%的生活用水和40%的灌溉需求。 目前， ...

科技日报大连1月22日电 （记者张蕴）22日，在中国特种设备检测研究院、中国科学院沈阳自动化研究 所、大连市科学技术局主办的首届特种机器人科产融合大会上，"特种机器人百人会"正式发起成立。此 举标志着我国特种机器人产业进入深度融合的新阶段。 "特种机器人百人会"首批成员由中国特种设备检测研究院和中国科学院沈阳自动化研究所联合召集，聚 焦"政产学研金服用"深度融合，旨在构建特种机器人产业协同创新生态，破解特种机器人产业发展瓶 颈，助力产业高质量发展。 特种机器人是推动石油、化工、电力、应急、消防等重点领域转型升级的关键装备，正逐步从技术验证 期转入场景深耕期。中国工程院院士陈学东表示："推动特种机器人产业高质量发展，既要聚焦核心技 术攻关，着力突破智能控制、精密传感、自主导航等技术，打造更可靠、更适配的装备体系；也要强化 科产融合，让科研成果真正对接市场需求，通过产学研用协同创新，加速技术转化与场景落地；更要注 重生态构建，汇聚各方智慧，凝聚发展合力，共同绘制产业发展蓝图。" 中国特种设备检测研究院党委书记、院长王军介绍，如今，特种机器人深度渗透到石油石化、应急消 防、核电运维等13大关键领域，成为保障国计民生 ...

Group 1 - The core viewpoint highlights the declining profitability of major South Korean battery companies, with all three—LG Energy Solution, Samsung SDI, and SK On—projecting operating losses by Q4 2025 [1] - LG Energy Solution is expected to incur a loss of 1.22 trillion KRW in Q4 2025, which could increase to 4.548 trillion KRW without U.S. tax incentives [1] - Samsung SDI anticipates a loss of approximately 300 billion KRW, while SK On expects a loss of 200 billion KRW in the same period [1] Group 2 - In response to these challenges, companies are shifting some production capacity towards the energy storage sector [2] - LG Energy Solution plans to invest 1.4 billion USD to convert its facility in Holland, Michigan, into a dedicated energy storage system production base, aiming to expand its energy storage battery capacity to over 50 GWh by 2026 [2] - Samsung SDI intends to convert part of its electric vehicle production lines in Indiana to produce energy storage batteries, targeting an annual production capacity of 30 GWh by the end of 2027 [2] - SK On will establish a new production line in South Korea with an annual capacity of 3 GWh for lithium iron phosphate batteries for energy storage [2]

在后续与中国团队合作开展的试验中，5名中风患者和10名健康受试者默念短语后，通过两次点头即可 触发设备将短语扩展为完整句子。例如，"我们去医院"可被扩展为"虽然时间有点晚了，但我不太舒 服，我们现在能去医院吗？"设备能根据心率升高及深夜时间等情境，自动补充语气与内容。 设备采集的信号由两个AI模块处理：一个负责从无声口型中识别词语；另一个则结合情绪状态与上下 文信息（如时间、天气等），将简短的词语扩展为完整且符合语境的表达。 在一项针对5名构音障碍患者（中风后常见言语障碍）的小型试验中，该设备在词语识别上的错误率为 4.2%，句子重建错误率仅为2.9%。与现有辅助语音技术（如逐字输入、眼动追踪或脑部植入）相比， Revoice能够实时、连贯地进行语句转换，大幅提升交流效率。 该设备不仅适用于中风康复，未来也可能用于支持帕金森病、运动神经元疾病等患者的交流需求。 科学家开发出一款能解码"无声之言"的项圈，名为"Revoice"。这款设备结合了高灵敏度传感器与人工 智能（AI）技术，佩戴舒适且可水洗，能够帮助中风患者恢复自然流畅的交流能力，而无需进行侵入 性脑部植入手术。相关研究成果发表于最新《自然·通讯》期刊。 ...

日前，北京智谱华章科技股份有限公司通过其官方公众号发布GLM Coding Plan限售公告。公告指出， 随着GLM-4.7系列模型上线，用户数量迅速增长，导致算力资源出现阶段性紧张。 这是AI产业算力吃紧的一个缩影。随着GPT、DeepSeek等大模型的算法突破和应用普及，算力需求水涨 船高。数据显示，我国AI芯片市场规模预计2028年将超一万亿元，约占全球市场的30%。面对庞大的市 场需求，自主可控的高质量AI算力供给已成为我国抢占人工智能产业应用制高点、全方位赋能千行百 业的前提条件。 算力缺口从何而来？国产AI算力又当如何破局？记者就此采访了有关专家。 同时，我国算力资源分散，存在"碎片化"问题。各服务商算力资源接口和协议不统一，跨区域跨主体算 力调度能力较弱，导致算力资源利用率偏低。此外，产业发展制度环境仍待完善，数据确权、使用和交 易等方面的规则有待细化，企业标准与合规方面的挑战日益凸显。 与此同时，人工智能正在加速落地千行百业，带来算力需求激增。目前，全国已落地的算力应用项目超 过1.3万个，建成的各级智能工厂超过3万家，并覆盖工业、金融、交通、医疗、教育等重点行业。"随 着我国人工智能应用加速 ...

Group 1 - The 2026 Leiden World University Ranking highlights Zhejiang University at the top, with Harvard University dropping to third place, and eight Chinese universities in the top ten, indicating China's strong position in global academic research [1] - The ranking is based on bibliometric data, specifically paper output and citation counts, which are considered authoritative measures of academic impact [1] - The rise of Chinese universities reflects a significant achievement, but it is essential to recognize that paper output and citation counts are just one dimension of evaluating research strength [1] Group 2 - The historical shifts in the center of scientific innovation, from Italy in the 17th century to the United States in the 20th century, illustrate that multiple innovative factors contribute to maintaining research advantages [1] - The release of the ranking serves as both encouragement and a reminder that the ultimate strength of a technological power lies in talent reserves, an open innovation ecosystem, and a continuous scientific spirit [1] - The discussion around AI and the need for innovative thinkers emphasizes the importance of designing talent selection systems that support diverse intellectual growth and the efficient transformation of research outcomes into practical applications [2]

Core Insights - The research team from Peking University has developed a new analog computing chip for non-negative matrix factorization, significantly improving processing speed and energy efficiency compared to current digital chips [1][2] Group 1: Technology Overview - Non-negative matrix factorization (NMF) is a powerful data dimensionality reduction technique used in various fields such as recommendation systems, bioinformatics, and image processing [1] - Traditional digital hardware struggles with real-time processing demands due to computational complexity and memory bottlenecks when handling large-scale datasets [1] Group 2: Chip Performance - The new chip, based on resistive random-access memory (RRAM), achieves approximately 12 times faster computation speed and over 228 times better energy efficiency compared to advanced digital chips [1][2] - In image compression tasks, the chip maintains image quality while reducing storage space by half, and in recommendation system applications, it shows prediction error rates comparable to digital chip results [2] - In the MovieLens 100k dataset training task, the analog calculator achieved a speed improvement of 212 times and an energy efficiency improvement of 46,000 times compared to mainstream programmable digital hardware [2] Group 3: Implications for Industry - This research opens new pathways for real-time solutions to constrained optimization problems like non-negative matrix factorization, showcasing the potential of analog computing in handling complex real-world data [2] - The advancements could lead to innovations in real-time recommendation systems, high-definition image processing, and genetic data analysis, contributing to more efficient and lower-power artificial intelligence applications [2]