Chemistry is a discipline that studies the composition, structure, properties, and laws of change of substances. Its development history is the history of humanity breaking through natural phenomena and revealing the essence of matter. From the ancient alchemists' obsession with "turning stones into gold" to modern scientists manipulating atoms to synthesize new substances, driven by curiosity, humanity is constantly advancing towards the unknown. In recent years, artificial intelligence (AI) technology has developed rapidly and entered various industries. What kind of 'chemical reactions' will arise when AI encounters chemistry? In the field of chemistry, AI can process massive amounts of data, discover complex structures that are difficult for humans to detect, accelerate scientific research processes, and help solve complex scientific problems. From data acquisition to pattern discovery, and then to technological applications, AI technology is fundamentally reshaping chemical research, bringing infinite possibilities for humanity to explore the unknown. In recent years, traditional chemistry and chemical research methods have become inadequate in the face of complex chemical molecular structures and massive experimental data. ”Professor Wang Dan from Beijing University of Chemical Technology introduced to reporters that AI can simulate the human learning process, extract useful information from large amounts of data, optimize decision-making processes, and has shown great potential in the field of chemistry research. For example, Shanghai Jiao Tong University has utilized the Magnolia scientific model to accelerate the entire process of organic synthesis through a chemical language model for the first time. This large model has demonstrated outstanding capabilities in multiple benchmark tasks such as single/multi-step inverse synthesis, yield prediction, selectivity prediction, and reaction optimization. It has great potential in accelerating the discovery of real chemistry and is expected to solve the problem of repeated trial and error in experimental science. It provides a new research paradigm and method for accelerating organic chemistry synthesis with large language models. In the interdisciplinary field of chemistry and other disciplines, AI also has great potential. We are mainly engaged in cutting-edge science of supergravity and engineering research of nanomaterial products. Currently, our research group is collaborating with experts in the computer field to jointly develop AI intelligent agent prototype systems for specific organic-inorganic nanocomposite material systems. ”Wang Dan said that the research team is currently using deep learning algorithms to construct formula models. The experimental results show that under the constraints of given performance indicators, cost, and environmental impact, the AI system can quickly generate material formulas that meet the requirements. I think the most significant improvement brought by AI is mainly reflected in shortening the research and development cycle, reducing research and development costs, and increasing the probability of discovery. ”Dong Xufeng, Vice Dean of the School of Materials Science and Engineering at Dalian University of Technology, said that AI can assist scientists in multiple research steps such as new material design, performance prediction, and process optimization. For example, for biomedical materials, AI can assist in designing the structure of porous scaffolds, predicting their mechanical properties and biocompatibility. AI can also analyze massive amounts of process parameters and final product performance data, identify the optimal preparation process window, and achieve precise control of the process. The quality of chemical experimental data, commonly known as the "thirst for data," directly affects the predictive and decision-making abilities of AI. However, currently, chemical AI is generally facing a 'thirst for data'. Wang Dan gave an example that the research and development of new chemical materials involves multi-source heterogeneous data such as experiments, simulations, and literature. However, this field has long been plagued by problems such as data fragmentation, semantic gaps, and insufficient cross scale correlations. High quality data is the foundation for training effective AI models, but obtaining, organizing, and standardizing chemical data remains a major challenge. He suggested building a unified high-quality data and knowledge system to achieve structured integration and dynamic updating of material characteristics and performance information. Dong Xufeng also believes that data scarcity and low data quality are important obstacles to the deep application of AI technology. Especially in the field of biomedical materials, the amount of data related to in vivo experiments and clinical research is not only small, but also has extremely high acquisition costs, high noise, and inconsistent standards. He suggested that data standardization and sharing should be promoted, and a standard format and sharing platform for material data should be established; Developing small sample learning and zero sample learning techniques to enable AI to learn how to draw lessons from one example to another; Integrate multi-source data, including simulated data, literature data, experimental data, and even failed experimental data. In addition, industry experts generally believe that data-driven AI still faces problems such as insufficient model representativeness and uncertain interpretability. To truly form practical technical methods, it is necessary to conduct systematic and in-depth research and exploration. As scientists, we not only need to know what it is, but also why, "said Dong Xufeng. Wang Dan believes that promoting the deep integration of AI and chemical research requires not only continuous optimization at the technical level, but also integration and integration at the thinking level. He analyzed that the current mainstream AI models mostly use general algorithms and lack adaptability to chemistry majors, resulting in a "language barrier" between AI and chemistry. It is not difficult to imagine that a PhD in chemistry and a PhD in computer science have unconsciously formed thinking patterns based on the characteristics of their respective disciplines during their education and research training. Thinking patterns are the core framework of human cognition and decision-making, and different thinking patterns determine different ways in which people understand the world, solve problems, and create value. Wang Dan believes that to promote interdisciplinary research on "AI+chemistry", it is necessary to gather experts from these two or more fields. Therefore, it is crucial to cultivate and select innovative talents who truly possess a composite knowledge system and interdisciplinary integration ability. As AI technology gradually matures, where is the boundary between its division of labor and that of human scientists? In Dong Xufeng's view, even if AI develops to a very high level in the future, at least in material research, AI will still face insurmountable boundaries, and the core position of scientists will not change. AI excels at optimizing and exploring within given goals and frameworks, but it cannot spontaneously propose a completely new and disruptive scientific problem. AI is a tool, it has no values. Researching what and why involves various aspects such as social needs and ethics. ”Dong Xufeng said that true scientific breakthroughs often come from connecting two seemingly unrelated fields. The ability to associate from a distance and the intuition and aesthetics based on profound knowledge are still the "privilege" of humanity at present. AI may become an incredibly powerful research assistant in the future, capable of handling all tedious, complex, and time-consuming calculations and data analysis work, freeing scientists from repetitive labor. But the 'brain' and 'soul' of research, namely the wisdom to ask questions, the responsibility to define directions, the inspiration to creatively integrate, and the responsibility to bear ethical consequences, will always belong to human scientists. ”Dong Xufeng believes that the future development direction of "AI+chemistry" should be a "scientist AI" symbiosis of human-machine collaboration and each showcasing their strengths, rather than mutual substitution. (New Society)