Abstract: In the face of complex and uncertain scientific research activities, the wide application of generative AI tools is leading to a new scientific research paradigm – “AI for Science”. By inquiring into the scientific research paradigm driven by artificial intelligence, this paper discusses the difference between the scientific research paradigm here and Kuhn’s “paradigm”. According to the current discussion of AI-driven scientific research paradigms, most of the findings are based on the evolution of Gray’s scientific research paradigms. But when we delve into the practice and application of the AI-driven scientific research paradigm, we find that it may not be evolving along the lines that Gray delineates. In addition, Gray’s division is mainly based on the change of scientific research means or methods, but according to the view of scientific practicalism, scientific research activities are composed of three main influencing factors: scientific research subject, scientific research means and scientific research object. Therefore, Gray’ s division has certain limitations, and it is necessary to further clarify the limitations of this classical division. Then, it further combs the existing historical forms of scientific research paradigm. After sorting out the different historical forms of the four groups of scientific research paradigms, it can be found that the development process from simple science to complex science, and the transformation process from philosophy to natural science, all the elements of scientific research activities have undergone transformative changes. But between physics and chemistry, only scientific methods have changed. On the other hand, natural history and mathematical experimental science have changed the means and objects of scientific research. When reviewing these different scientific traditions in history, it is found that the division of the historical form of scientific research is varied, and different subject positioning, vision and concern will form different division results. Whether it is the change of scientific tradition in a grand perspective or the change of scientific model in a specific field, it provides historical and forward-looking experience for the division of scientific research paradigm in the era of artificial intelligence. The rise of AI for Science has brought about the thinking of the revolution of scientific research paradigm, which provides a new perspective for the division of scientific research paradigm based on the uniqueness of the change of scientific research subjects. Based on this, two conclusions can be drawn. Firstly, the division of scientific research paradigms has many forms in history, which cannot be confined to the Gray’s division. Second, when sorting out the different historical traditions of scientific research paradigms, it is found that compared with traditional scientific research activities, there are some great differences in research subjects and other aspects of AI-driven scientific research activities, which is a significant change compared with previous scientific research paradigms. Therefore, the scientific research paradigm driven by artificial intelligence can be regarded as a new scientific research paradigm. Therefore, the truly revolutionary division is not only based on the change of scientific research tools, but also on the organic change of various elements in scientific research activities, the most important of which is the change of scientific research subjects, of course, there are also changes in scientific research objects, and artificial intelligence from Artificial Intelligence Generated Content (AIGC) to Artificial Intelligence Generated Science (AIGS). In the future, with the extensive and deep application of artificial intelligence in scientific research activities, it will likely contribute to a major revolution in the paradigm shift of human scientific research, and it will also more closely interpret the depth and breadth of scientific problems and social phenomena.