Core Viewpoint - The article emphasizes the emergence of "AI for Science" (scientific intelligence) as a transformative force in research, moving beyond traditional methods to integrate AI deeply into scientific discovery processes [1][2]. Group 1: Company Overview - SiLang Technology, founded in 2016 and rooted in the Chinese Academy of Sciences, has become a leader in the scientific intelligence sector, focusing on high-performance computing chip design [4]. - The company has completed seven rounds of financing, attracting investments from major industry players like CATL and Kweichow Moutai, as well as notable financial investors, ensuring comprehensive coverage of the industry chain [4]. Group 2: Technological Innovation - SiLang Technology has developed the MaPU (Matrix Processing Unit) architecture, which is a key innovation that allows for significant advancements in scientific computing [5][7]. - The MaPU architecture achieves over 90% core utilization in scientific tasks and is reported to be dozens of times faster than traditional CPUs, overcoming previous performance limitations [7]. Group 3: Product Development - The company has launched two core products: the "Tianqiong" supercomputer and the UCP baseband chip, which together form a dual-driven strategy to convert foundational technology into high-value solutions [8]. - The "Tianqiong" supercomputer, as China's first fully self-developed 3D scientific computer, significantly enhances computational speed and reduces communication latency, outperforming traditional supercomputers by 2 to 4 orders of magnitude [9][11]. Group 4: Application and Impact - The "Tianqiong" supercomputer has been instrumental in various fields, including biomedicine and new materials, facilitating drug development and addressing challenges in lithium battery research [13]. - The UCP baseband chip has been successfully applied in 5G and satellite communication, establishing itself as a preferred solution among domestic low-orbit satellite internet constellations [14]. Group 5: Strategic Positioning - SiLang Technology aims to build a complete scientific intelligence ecosystem that integrates architecture, computing power, data, and applications, driven by a strong team combining technical and commercial expertise [16]. - The company is positioned to compete globally in the scientific intelligence arena, with a focus on creating a "scientific data factory" to ensure high-quality data generation and model effectiveness [16][18].

Core Viewpoint - The "Panshi" scientific foundational model, developed by a team from the Chinese Academy of Sciences, was unveiled at the 2025 World Artificial Intelligence Conference in Shanghai, aimed at assisting scientists in interdisciplinary research [1] Group 1: Model Capabilities - The "Panshi" model features core capabilities such as scientific literature extraction and integration, scientific knowledge representation and reasoning, and scientific tool orchestration and planning [1] - It enables a complete research process from hypothesis generation to experimental validation and discovery of laws, functioning as a cross-disciplinary research support platform combining AI and science [1] Group 2: Customization and Development - The model is deeply customized for the scientific field based on domestic open-source large models, mastering core theorems, laws, and professional knowledge across six major disciplines: mathematics, physics, chemistry, biology, earth science, and astronomy [1] - Two scientific intelligent entities, "Literature Compass" and "Tool Scheduling Platform," have been developed to assist researchers in various tasks [1] Group 3: Application Examples - In mechanics research, the model demonstrates strong capabilities in understanding and predicting scientific data, efficiently calculating surface pressure fields of high-speed train models in various fluid environments to support design [1] - In high-energy physics research, the model can automatically decompose and efficiently plan particle physics research tasks, significantly enhancing particle simulation speed and reconstruction efficiency, aiding in the exploration of the fundamental composition of matter and the basic laws of the universe [1]