AI As Scientist? Machine-Written Papers Clear Academic Reviews, Raise Questions

An artificial intelligence has independently conceived, written and submitted a scientific paper that passed peer review at a major academic conference, marking what researchers are calling a turning point in the history of science. The system, known as The AI Scientist, was developed by Sakana AI, a Tokyo-based research company, and is the first of its kind to automate nearly the entire scientific process without meaningful human involvement. 

Its creators describe it as capable of generating research ideas, running experiments, writing code, analysing results and producing a fully formatted manuscript, all on its own. The findings were published in the journal Nature.

How does it work?

The AI Scientist follows a workflow that closely mirrors that of a human researcher. It begins by scanning existing literature to generate original research directions, filtering out topics that have already been well explored. It then writes and executes its own experimental code, drafts the paper in the academic formatting language LaTeX, and even selects relevant studies to cite.

In a final step that underscores just how autonomous the system is, it submits its work to a separate AI reviewer, trained to assess papers on accuracy, quality and originality, before the manuscript goes anywhere near a human.

The test

To put the technology to a real-world test, the Sakana team submitted three AI-generated papers to a workshop at the 2025 International Conference on Learning Representations (ICLR). Human reviewers were informed that some submissions might be machine-written, but were not told which ones.

One of the three papers was accepted. It received scores of 6, 7 and 6 from reviewers, a modest but meaningful result. The paper was subsequently withdrawn, in line with a transparency agreement made with conference organisers ahead of the experiment.

"This achievement demonstrates the growing capacity of AI for making scientific contributions," the Sakana team wrote, describing it as "a potential paradigm shift in how research is conducted."

Cause for caution

The researchers are candid about significant shortcomings. The system is prone to hallucinations - fabricating references to papers that do not exist, or reproducing the same figures across different sections of a manuscript.

Ethical concerns loom larger still. A flood of AI-generated submissions could overwhelm already-stretched peer-review systems, making it harder for the work of human scientists to receive the attention it deserves. There are also worries the technology could be misused to conduct unethical research or artificially inflate the publication records of individual researchers.

Sakana has called for clear disclosure standards to be established as the technology develops, arguing that transparency will be essential to preserving the integrity of the scientific process.