Artificial intelligence-generated disproportionate rat genitalia makes its way into a peer-reviewed journal

Editors of the magazine Frontiers in cell and developmental biology They retracted a paper after readers pointed out to them that the supporting images had been incorrectly generated by an AI-powered image generator. In their retraction, the editors stated that the reason for the retraction was that “concerns have been raised about the nature of the numbers generated by artificial intelligence.”

In the article, which included research on stem cells in small mammals, the authors included images that clearly depicted the anatomy of mice generated by the AI ​​system. In one photo, one mouse appeared to have a penis and testicles larger than the rest of its body, something that does not occur in nature. Also, some of the accompanying texts were incomprehensible. Another image showed a mouse cell that did not resemble the actual structure of a mouse cell.

The paper's incongruous images are likely to add to ongoing discussions in the scientific community about using artificial intelligence to generate text or images for use in technical research, especially those published in accredited journals.

In this case, it is not clear how such problematic images ended up being published in a peer-reviewed journal. The authors, a joint team from Honghui Hospital and Jiaotong University in China, did not try to hide the fact that they used artificial intelligence to create the images; They even credited Midjourney.

Some journalists noticed this the border It has a policy that allows the use of AI-generated material as long as its use is disclosed, which was the case in this case. But the policy also suggests that attempts should be made to validate anything produced by such systems, which was clearly not the case in this mix-up.

Editors in the border He first posted a note on the newspaper claiming that the article had been corrected and that a new version would be published shortly. Not long after, the paper was withdrawn.

The errors made by the paper's authors and the team at the journal that approved its publication are likely to be the first of many to come, although it remains unclear what changes will be needed to prevent such errors from occurring in the future. .