Sakana AI unveiled a landmark achievement this week: an AI-generated scientific paper that passed peer review in a top-tier journal. The paper was authored entirely by AI Scientist v2, Sakana's research-conducting AI system, and accepted by Nature in record time. The system took just 15 hours to generate the paper and spent approximately 40 in compute costs to do so.
The paper represents a watershed moment for AI in scientific research. Rather than simply analyzing existing data, AI Scientist v2 conducted original experiments, designed novel hypotheses, and produced findings sufficiently rigorous to convince peer reviewers at one of the world's most prestigious journals. The work challenges assumptions about AI limitations in creative scientific reasoning and hypothesis generation.
"Some of the AIs ideas seemed truly creative, yet the system struggled with execution."
The paper scored 6.33 on average from peer reviewers, placing it in the top 55% of papers submitted to Nature. This performance exceeded the average rating for human-authored papers, though reviewers noted that AI Scientist's creative problem-solving sometimes outpaced its ability to execute methodologically. The finding suggests AI may be better suited for certain aspects of research—hypothesis generation and experimental design—than others.
The implications ripple across academia and industry. AI systems that can conduct research, write papers, and pass peer review compress the time required for scientific advancement by orders of magnitude. For researchers using AI as a collaborator rather than a replacement, the technology offers the potential to dramatically accelerate discovery timelines. However, questions remain about reproducibility, methodological rigor, and whether AI papers might systematically bias journals toward certain types of findings.