Artificial intelligence is helping humans gain a deeper understanding of the moon. During the 2024 China International Big Data Industry Expo, the Institute of Geochemistry of the Chinese Academy of Sciences and Alibaba Cloud jointly released the world's first "multi-modal professional model of lunar science" on August 29. Generative AI represented by large models has brought new opportunities for lunar science research. A viewer is watching an introduction to a multimodal professional model of lunar science. The reporter learned from the press conference that with the rapid development of deep space exploration technology, lunar exploration data has shown an explosive growth. "At present, there are more than 2 PB of lunar and planetary exploration data around the world." Liu Jianzhong, member of the Party Committee and researcher of the Institute of Geochemistry, Chinese Academy of Sciences, said that there are numerous impact craters on the moon, which have an important impact on the evolution of the moon itself. However, due to the large number of impact craters and their complex shapes, research is very complicated. The most commonly counted impact craters now are those with a diameter greater than one kilometer, and the latest statistics show over one million Liu Jianzhong said that as the diameter of the impact crater decreases, its number increases exponentially. It is impossible to identify or count every type of impact crater solely relying on manual labor. The rapidly developing artificial intelligence models today provide a solution to this problem. Zeng Zhenyu, Vice President of Alibaba Cloud Intelligent Group and General Manager of Solution R&D Department, told reporters that since last year, Alibaba Cloud has fully communicated with the Institute of Geochemistry of the Chinese Academy of Sciences, which took nearly a year to polish and expand the multi-modal large model to the field of lunar research. The meteorite impact crater is the first pilot project of our cooperation. Nowadays, in terms of age and morphology discrimination of lunar impact craters, the accuracy of multimodal professional large models in lunar science can reach over 80%, exceeding our expectations Zeng Zhenyu said that with the deepening of cooperation between the two sides, in the future, multimodal large-scale models can be expanded to analyze and study more geological landforms on the moon. By utilizing large-scale models, the efficiency of identifying impact craters can be significantly improved and research costs can be reduced Liu Jianzhong stated that in this project, the Institute of Geochemistry provided a lot of textbooks and literature on the moon and planets, nearly 20000 PDF documents, and about 50 TB. After processing these documents, Alibaba Cloud became the knowledge center for training "big models". The professional large-scale model of the moon will greatly accelerate the processing of massive data and help researchers discover new scientific discoveries Ouyang Ziyuan, an academician of the CAS Member and the first chief scientist of China's lunar exploration project, commented. (New Society)