Microsoft(US:MSFT) 量子位·2025-05-20 03:32Core Viewpoint - Microsoft has launched Microsoft Discovery, an enterprise-level AI research platform that enables scientists and engineers to discover new materials in significantly reduced timeframes, exemplified by the identification of a new cooling agent in just 200 hours instead of months or years [1][2][6][12]. Group 1: Microsoft Discovery Overview - Microsoft Discovery combines professional AI agents with high-performance computing to bridge the gap between domain experts and computational capabilities [3][4]. - The platform aims to accelerate scientific outcomes with greater speed, scale, and precision [5]. - It allows researchers without programming backgrounds to utilize supercomputing capabilities through natural language interactions [14][16]. Group 2: Application and Impact - The platform successfully identified 367,000 potential alternatives to harmful PFAS cooling agents, showcasing its practical application in real-world scenarios [9][10]. - Microsoft Discovery is being applied across various fields, including pharmaceuticals, chip design, and engineering manufacturing [15]. - The platform significantly lowers the barrier to advanced research, enabling domain experts to conduct experiments without needing to hire specialized programming personnel [17]. Group 3: Technical Components - Microsoft Discovery integrates two key components: a foundational model for planning and specialized models trained for specific scientific fields [19]. - It employs a knowledge graph engine to connect proprietary data with external research, enhancing understanding of complex scientific data [22]. - The platform supports collaborative reasoning and iterative research cycles, ensuring transparency and expert involvement in the research process [25][26]. Group 4: Future Prospects - Microsoft Discovery is built on Azure infrastructure, allowing small laboratories to access the platform with costs based on cloud service models [30]. - The architecture is designed to potentially integrate quantum computing capabilities in the future, expanding its application in scientific research [31][32].