ScienceOne, AI-Powered Research Platform, Launches

On April 29th, the Institute of Automation at the Chinese Academy of Sciences (CAS) unveiled ScienceOne, an intelligent research platform based on scientific foundation models, at the 8th Digital China Construction Summit. The platform aims to accelerate scientific discoveries through advanced AI technology, fostering interdisciplinary collaboration and a new paradigm of multidisciplinary research. During the launch event, Dr. Zeng Dajun, Deputy Director of CAS's Institute of Automation, introduced two primary products: S1-Literature, a literature assistant, and S1-ToolChain, a scientific tool scheduling platform. These tools leverage the institute’s strengths and those of its partners to tackle common issues in general large models, such as hallucination, lack of specialized knowledge, and weak logical reasoning, specifically in the scientific domain. S1-Literature Literature Assistant S1-Literature, co-developed with CAS’s National Science Library, utilizes the largest domestic scientific literature database and real-time open-source materials. It excels in generating high-quality literature reviews with minimal user input. For instance, a simple command like "write a review on [specific topic]" can produce a detailed, structured overview of thousands of papers. Additionally, S1-Literature offers multiple auxiliary tools, including mind maps, citation tracing, research graphs, and technological pathway extraction, to aid researchers in comprehensively understanding and critically reading papers. The tool currently supports disciplines such as mathematics, physics, and materials science, with plans to expand to all fields in the future. S1-ToolChain Scientific Tool Scheduler S1-ToolChain focuses on enabling autonomous cross-disciplinary data interpretation, scientific computation, and simulation. Through a standardized scientific model protocol, it can integrate various general and specialized scientific models and tools, orchestrating them as intelligent agents to perform complex task flows. Currently, the scheduler has integrated nearly 300 tools from multiple fields, including data analysis, differential equation solving, discrete optimization, and multiscale simulations. Researchers can initiate sophisticated processes with straightforward commands, such as “complete a protein sequence,” which triggers a series of operations from intent recognition to task execution, forming a closed-loop workflow. Potential for Accelerating Research Innovation The integration of AI into scientific research has already delivered groundbreaking advancements in fields like biology, materials science, and pharmaceuticals. This signifies a shift towards a new paradigm of multidisciplinary research. The Institute of Automation at CAS has capitalized on this trend by leveraging its well-established disciplinary system and top-tier multidisciplinary research teams. By offering robust technical support and versatile applications through the ScienceOne platform, the institute supports scientists across various domains and aims to drive the development of next-generation research paradigms. Future Plans The development team plans to open-source the scientific foundation model S1-Base and launch a scientific intelligence agent factory S1-Agent. These initiatives will further enhance the platform's functionality and toolset, contributing to the advancement of smart research practices. Industry Reactions ScienceOne has garnered significant attention and positive feedback from the scientific community. Experts commend it for filling gaps in existing research tools and substantially improving research efficiency and quality. They predict it will become a crucial infrastructure for future scientific endeavors. By streamlining data processing and tool deployment, ScienceOne enables researchers to focus more on core scientific questions. Institute Profile Founded in 1956, the Institute of Automation at CAS is a leading national research institution with extensive expertise in robotics, intelligent information processing, control theory, and complex systems. Over the years, the institute has consistently contributed to the nation's technological innovation and advancement. The launch of ScienceOne underscores the institute's commitment to pushing the boundaries of AI in scientific research and positioning itself at the forefront of future innovation. The release of ScienceOne marks a pivotal moment in the AI4Science movement. It represents a significant step towards smarter and more efficient research practices, reducing the time researchers spend on data management and allowing them to concentrate on critical scientific challenges. With the continued development and expansion of its tools and models, ScienceOne is expected to play a transformative role in the scientific community, heralding a new era of intelligent research.