Is Human-Free Academic Research Possible? The First “AI Scientist” Completes the Paper Process on Its Own
A new study published in the journal Nature has revealed a revolutionary development in the scientific world. A team consisting of researchers from the University of British Columbia, the University of Oxford, and Sakana AI introduced “The AI Scientist’’ capable of conducting end-to-end research without human intervention. This work has entered the literature as the first example of artificial intelligence executing the entire scientific research process on its own.
The system in question can autonomously manage all stages of an academic research. The AI first generates original project ideas and conducts a literature search to verify the novelty of these ideas. Then, it writes the necessary code for the experiments, debugs, and analyzes the data to generate figures. At the end of the process, it writes up all the obtained findings in a full academic paper format.
When the system was tested in the real world, striking results were obtained. The researchers submitted three papers produced by the AI to the I Can’t Believe It’s Not Better (ICBINB) workshop of the International Conference on Learning Representations (ICLR 2025), with the full approval of the conference leadership and workshop organizers. Although the reviewers were aware that some of the submissions were AI-generated, they did not know which ones. In this blind review process, one of the papers exceeded the workshop’s acceptance threshold and successfully passed peer review. However, researchers note that the system is not yet perfect. It is highlighted that the AI can occasionally make incorrect citations (hallucination) or present insufficiently mature ideas. Furthermore, the system can currently only conduct research in the field of computer science.
This autonomous technology is expected to accelerate the pace of scientific discovery to an unprecedented level in the future. The project team envisions that in the coming years, communities of AI could be established that not only discover new information but also use these discoveries to improve themselves. According to experts, it is emphasized that if developed with the necessary oversight and principles, this system could be a turning point capable of transforming the way scientific research is conducted globally and the future of academia. However, the extent to which a change of this scale will alter the validity criteria of scientific studies and academic acceptance conditions emerges as a topic of debate.