A recent study published in the journal Nature shows that a system called "AI Scientist" has been able to complete the entire process from proposing research questions to writing papers, and even peer review, almost without human intervention. Some papers have even passed international academic conference reviews. This system was developed by Sakana AI, a Tokyo based artificial intelligence company in Japan, and its workflow simulates the scientific research process of human scientists. Firstly, AI will propose multiple research directions and screen through literature databases to eliminate those topics that have already been studied. Subsequently, it will conduct experiments by calling existing templates or autonomously generating code. After the experiment is completed, the AI will use LaTeX (document formatting system) to draft a formal scientific paper. In addition to standard paper content such as writing methods, results, and discussions, it can also search for relevant research and cite it. Finally, the system will call an automated review program, which is another trained AI model capable of scoring papers based on their accuracy, quality, and originality. The team will submit three fully AI generated papers to a machine learning seminar at the International Conference on Learning Representation (ICLR) in 2025. Although the reviewing experts were informed in advance that some papers may be generated by AI, they do not know which ones specifically. In the end, one of the papers received scores of 6, 7, and 6, with an average score of 6.33. It successfully passed the review and was accepted. According to the transparency agreement of the experiment, this paper was subsequently voluntarily withdrawn. The team stated that this result indicates that AI has begun to have the ability to independently complete some scientific research work, which may have a profound impact on the organization of scientific research in the future. However, they also emphasized that the system still has obvious limitations, such as the possibility of "illusion" issues, including citing non-existent papers or reusing charts. Research suggests that a large number of AI generated papers may increase the burden of peer review, affect the normal operation of the scientific research evaluation system, and even be used by individual researchers to improperly increase the number of academic achievements. (New Society)