Wireless AI, that is, the deep integration of AI and communication technology, is one of the six major demand visions of 6G (IMT-2030) announced by ITU, and has become the focus of research and application in the global academic and industrial circles. This report returns to the first principles, starting from the background and key technical characteristics of wireless AI research, sorting out the progress of cutting-edge research at home and abroad, and discussing the current research challenges and difficulties of wireless AI. In response to the pain points of wireless AI datasets, combined with digital twin technology, this report focuses on the detailed introduction of the use of 3D channel modeling theory and methods based on electromagnetic computing, AI technology to achieve real twins of channels and 3D scenes, and the twinning and construction of high-fidelity data for wireless AI scenes. Finally, based on the generated high-fidelity dataset, this report focuses on the application and thinking of wireless AI in the physical layer field.