Nature重磅：华人学者推出“AI机器人科学家”，自主做实验，仅用90天发现高性能催化剂

Core Insights - The article discusses the development of a multimodal AI robotic platform named CRESt, aimed at accelerating the discovery of advanced materials through real-world experimental validation [2][3][4]. Group 1: CRESt Platform Overview - CRESt integrates large multimodal models (LMM) with knowledge-assisted Bayesian optimization (KABO) and robotic automation to enhance material design and performance optimization [3][8]. - The platform utilizes a combination of chemical composition, text embeddings, and microscopic structural images to create a coherent framework for material exploration [7][8]. Group 2: Experimental Achievements - In a three-month period, CRESt autonomously explored over 900 catalyst chemical compositions and conducted 3,500 electrochemical tests, identifying a state-of-the-art catalyst with a cost-specific performance improvement of 9.3 times compared to traditional catalysts [4][12]. - The identified catalyst composition includes elements such as palladium, platinum, copper, and others, showcasing the platform's ability to navigate complex chemical spaces [12]. Group 3: Technological Innovations - CRESt's core innovation lies in its ability to interpret and synthesize heterogeneous data types, allowing for a multidimensional understanding of material systems [8][9]. - The platform's integration with advanced robotic synthesis and characterization enables a rapid feedback loop, transforming traditional material development into an agile closed-loop system [8][9]. Group 4: Future Implications - The framework of CRESt suggests a future where experimental laboratories operate as intelligent ecosystems, driven by AI to autonomously guide research directions and experimental designs [13][14]. - The success of CRESt in the field of electrochemical catalysis indicates its potential transformative impact across various scientific domains, paving the way for rapid prototyping in renewable energy, electronics, and pharmaceuticals [13][14].