Abstract: This article reviews milestone breakthroughs achieved by artificial intelligence (AI) in scientific research across various fields, including mathematics, physics, chemistry, biology, materials science, energy, pharmaceuticals, and integrated circuit design, as well as in the recursive self-evolution of AI itself (AI for AI). The significant progress of AI for Science (AI4S) suggests that using AI to first advance scientific and technological research, then leveraging those new scientific and technological achievements to drive economic and social development may yield greater benefits than directly applying current AI technologies to human production and daily life. The article also points out that the development of AI4S still faces great challenges.