Home - ActivitiesGenerative artificial intelligence (AI) is an important direction in the development of artificial general intelligence (AGI). It primarily involves designing AI algorithms to learn from multimodal, high-dimensional, and complex sample distributions and generate new samples. It forms the methodological foundation for current AI applications in areas such as automated question answering, cross-modal generation, and AI for science. The underlying foundation of AGI is mathematics and statistics. This talk mainly introduces the background, mathematical/statistical principles, and challenges of generative AI. It further introduces AI methods for constructing controllable/conditional generation based on optimal transportation, and applies them to problems such as medical image generation, multimodal image-text alignment, and molecular structure generation. Finally, it summarizes and forecasts the development and prospects of AGI.