机器之心发布 机器之心编辑部 SAIS Talk 是上智院主办的前沿技术分享会，迄今已成功举行 15 期，讲者背景多元，既有曾深度参与诺贝尔 奖评选的顶尖学者，也有活跃在科研一线的在读博士，以此激发灵感、共建生态。 9 月 26 日晚 ，五位来自共性技术、物质科学、生命科学、地球科学等方向的青年研究员将接连登场，分享 核心工作和创新思辨，内容涵盖 表征学习、催化反应预测、生物分子动态模拟、单细胞图谱、全球天气预 测 等多个领域。 当 AI 与科学深度融合，当年轻力量碰撞前沿课题，我们或将在不远的未来见证 "AI 爱因斯坦" 的诞生。诚 邀您与最具活力的 AI 力量同行，迎接科学发现的黄金时代的来临。 （赴 上海徐汇西岸的上智院 参与活动请发送姓名、机构、手机号至 sais@sais.org.cn ，报名截止 9 月 25 日 18:00，报名确认会以邮件在 21:00 之前回复） 活动议程 每个分享环节含 5-10 分钟交流 在全球人工智能浪潮奔涌向前的当下，创新的核心驱动力正越来越多地来自年轻一代。他们敢于挑战前 沿、不惧失败，正以跨界融合之姿重新定义科学发现的范式。 作为长期关注科学智能（AI for S ...

Group 1 - The 2025 Bund Conference in Shanghai showcases advancements in artificial intelligence across various sectors, featuring nearly 200 companies and over 30 new technology products [1] - Fudan University's Fuxi meteorological model provides global weather forecasts for the next 15 days, aiding agricultural production with precise data on temperature, wind speed, and precipitation [2] - AI is expected to significantly reshape society and drive industrial upgrades and economic structural changes, as stated by Zhang Hongjiang, a partner at Source Code Capital [2] Group 2 - AI is enhancing healthcare with applications like Fourier's "Smart Rehabilitation Port," which uses upper limb rehabilitation robots, and Ant Group's AI health application AQ, which mimics a doctor's diagnostic process [2] - Ant Group's CEO, Han Xinyi, emphasizes that AI in healthcare aims to provide personalized and accurate recommendations, although AI will not replace doctors in the foreseeable future [2] - Experts at the conference noted the deep integration of AI across various industries and sectors [2] Group 3 - The integration of AI and robotics is giving rise to the embodied intelligence industry, which involves three stages: perception, planning, and control [4] - Future AI is expected to learn through real-world interactions, evolving from a data-driven learner to an experiential explorer [4] - Challenges in embodied intelligence development include data collection and quality issues, as well as difficulties in aligning models with robotic control modalities [4] Group 4 - The current state of AI is likened to a desert with only a few opportunities, indicating a potential for explosive growth in the field [5] - The conference featured demonstrations of robots engaging in activities like boxing, showcasing the practical applications of AI and robotics [5]

Core Insights - Shanghai has accumulated 82 large models that have passed the filing process, indicating a significant advancement in AI development in the region [1] Group 1: AI Development Initiatives - The Shanghai Municipal Economic and Information Commission is focused on creating a demonstration area for vertical applications that empower various industries [1] - The "Mold Shaping Shanghai" project is being implemented, concentrating on six key areas, enhancing the "AI+" initiative [1] - Vertical models are accelerating deployment, with the establishment of national AI demonstration application bases in manufacturing, finance, healthcare, and education [1] Group 2: Innovative AI Models - The "Fuxi" meteorological model and the "Morning Star" protein design model are highlighted as industry-leading innovations [1] - The company is accelerating the mass production of embodied intelligence, with the establishment of a humanoid robot innovation center and the release of the world's first embodied motion large model "Longyue" [1] - These advancements are aimed at empowering industrial manufacturing and logistics transportation scenarios [1]

Core Insights - The article discusses the breakthrough of Shanghai's "Fuyiao" meteorological model, which aims to provide highly localized weather forecasts for urban areas, enhancing precision and response time in weather prediction [1][2][3]. Group 1: Technological Advancements - The "Fuyiao" model has improved spatial resolution by three times to 1 kilometer and reduced inference time from 10 minutes to 3 seconds, likening it to a "digital nervous system" for cities [3]. - The model integrates multi-modal data from radar, satellites, and numerical forecasts, achieving an end-to-end forecasting loop for short-term severe weather events [4]. - The CFFF (Computing for the Future at Fudan) platform, with a computing power of 40 PFlop/s, supports various scientific fields, including life sciences and atmospheric sciences, enhancing research capabilities [7][12]. Group 2: Collaborative Research and Development - The Shanghai Intelligent Research Institute (上智院) collaborates with universities and industries to drive innovation across multiple scientific domains, focusing on AI-driven research paradigms [4][12]. - The CFFF platform has linked nearly 4,000 users across 51 departments, supporting numerous high-impact research papers and national projects [12]. - The institute aims to break down barriers in vertical scientific fields by developing specialized AI models, such as the "Fuxi" meteorological model and the "Nüwa" life sciences model [14][16]. Group 3: Practical Applications - The "Fuxi" meteorological model has demonstrated its practical value by accurately predicting the landing point of Typhoon "Bebinca," outperforming other forecasting models [14]. - The "Nüwa" model focuses on drug development, particularly in siRNA research, significantly improving the accuracy of drug efficacy predictions [16][18]. - The "Suiren" molecular model addresses core challenges in material science, aiding in the development of sustainable materials and enhancing battery research [18][19]. Group 4: Climate and Social Sciences - The Planet Intelligence@Climate (PI@Climate) initiative aims to create a climate science language model that integrates data from various disciplines to address complex climate issues [22][23]. - The initiative has developed a benchmark system to validate the capabilities of climate models, enhancing their applicability in real-world scenarios [23][25]. - The collaboration between AI technology and humanities research is exemplified by the "Chinese Civilization Model," which seeks to analyze human interactions and cultural heritage [29][30]. Group 5: Infrastructure and Ecosystem Development - The establishment of a scientific corpus platform has enabled the collection and management of over 10 PB of high-quality scientific data, facilitating breakthroughs across multiple disciplines [42]. - The integration of AI tools and data governance is crucial for enhancing research efficiency and ensuring data security [39][42]. - The collaborative efforts between the Shanghai Intelligent Research Institute and Fudan University aim to create a comprehensive ecosystem for scientific intelligence, focusing on the integration of computing power, algorithms, and data [35][36].