AI generates first peer-reviewed scientific paper in Japan

A new development in artificial intelligence (AI) has taken place in the field of science. A Japanese AI system called AI Scientist-v2 has independently created a peer-reviewed scientific paper. The paper, titled "Compositional Regularization: Unexpected Obstacles in Enhancing Neural Network Generalization," was accepted at a top machine learning conference, ICLR 2025. This achievement shows that AI can design experiments, analyze data, and write coherent research conclusions without human help. Reviewers did not know the paper was written by an AI and rated it highly, which highlights its capabilities. Some experts believe this could lead to AI playing a larger role in scientific discovery. However, not all scientists agree on the implications of this event. Yann LeCun, an AI expert, argues that current AI lacks true understanding and reasoning abilities. He believes AI has not yet reached the level of forming genuine mental models necessary for original discoveries. The AI system might have produced a research paper, but understanding the content remains a question. Sakana.AI, the company behind the AI, acknowledged the ethical dilemmas of their experiment and withdrew the paper from the conference. They see their work as an experiment rather than a final product. Currently, AI is being used to assist in research tasks, speeding up processes like literature reviews and experimental design, but it is still a tool within the scientific community. The paper produced by AI Scientist-v2 introduced new ideas but did not make groundbreaking findings. LeCun points out that true originality in science involves deeper intuition and insight. He believes AI still lacks the capability to understand and predict in ways humans do. As AI continues to improve, it may help scientists by generating new hypotheses and automating laboratory tasks. The journey of AI in science is ongoing. While we are not at a point where AI can fully understand and drive research, we may be getting closer. The development raises questions about how AI will change the scientific process in the coming years.