Generative AI's working principle can be likened to the development of an apprentice painter. It first analyzes vast amounts of data (such as millions of articles, paintings, or music) to learn the patterns and styles within them, much like a painter copying a master's work. For example, when fed the image "a sunset on a coast," it can combine its learned color combinations and compositional rules to generate a landscape painting that has never existed before. The core technologies underpinning this capability include generative adversarial networks (GANs) and the Transformer model. The former is like two artists competing against each other: one generates artwork, the other verifies authenticity, and the output quality improves through repeated competition. The latter, like a novelist skilled at understanding context, can generate coherent and long-form texts.