This paper briefly introduces the application of artificial intelligence in space science. In some simulated satellite tasks in space, artificial intelligence can provide assistance when combined with traditional methods to produce better results. In the realm of satellite sampling and image reconstruction endeavors, we employ Large Language Models (LLMs) to enhance the CUDA operator within the data simulation and image reconstruction phases. This strategic application has yielded a marked improvement in operational efficiency and has effectively addressed computational bottlenecks that have historically plagued conventional methodologies. In addition, we speculate that we can use diffusion models to simulate the reconstruction of images for each period. If feasible, after the actual satellite operation time reaches the upper limit, we can approximately obtain more period reconstruction results.