The picture you see above shows a Mouse with huge genitalswas generated by Middle Streeta popular generative artificial intelligence software that can create images from text input, and appeared in a scientific article published on February 13, 2024 in the journal Frontiers in Cell Biology and Development. The paper, authored by Chinese researchers Xinyu Guo, Liang Dong and Dingjun Hao, examines the connection between spermatogonial stem cells and a specific biochemical system that mediates cancer and infection.
But what interests us here is not the content of the article, but the numbers it contains, because they go directly into it. increasingly heated debate about the effects of a inappropriate use of AI and more generally to the problems of the system for examining scientific publications.
You don't have to be a Jedi Master in biology to understand that this is a number anything but precise from a scientific point of view. In fact, it seems obvious that the rat pictured has a penis. absolutely disproportionate compared to the rest of the body, even equipped with four testicles. Even when we read the writing on the figure, we immediately understand that something does not fit: testtomcels, bite And iollotte sserotgomar is me non-existent wordswhich means nothing.
All of this would mean little if this character wasn't part of one real scientific articlemainly subjected to the so-called process Peer review: Put simply, the content has been verified and validated – at least in theory – by independent academic experts. It is therefore not surprising that the article has been removed three days after the following publication many reports from other researchers about the scientific inconsistency of this number and the other two numbers included in the article.
But the question is this: How did these numbers get into a scientific article? Let's make one point clear right away: the magazine that published the article, like any other from the Frontières editorial group, does not prohibit the use of generative AI for image production, provided the use of such tools is declared and the images are checked by human review. The authors of the article correctly stated in the captions that the images were created using AI. However, it remains to be clarified how it is possible that they considered the inclusion appropriate “Meaningless” numbers in her article, and that the newspaper's critics apparently had nothing to complain about. Currently, he has not made any statement on the subject and has limited himself to this thanks for the reports.
Of course, there was a suspicion that, in addition to the images, the text of the article was also generated by artificial intelligence, but previous analyzes have confirmed this inconclusive.
The scope of this story will worrying Until we realize that if these images have passed the peer review process, who knows how many more realistic images have made it through without anyone noticing.
This episode is therefore a little like the child screaming: “The king is naked”: in fact, researchers have been complaining for years about the strong pressure on the academic publishing system Quantity of items produced, also with economic incentives. A real “race to publish” with tangible consequences for the quality of the articles themselves and for the careers of scientists. From this perspective, AI can represent a way to produce more in less time, and that sometimes happens You're going too far if you neglect it for errors and inconsistencies.
It is no coincidence that this was calculated in a study recently published in the prestigious journal Nature 2023 was the year with the record number of recalled items after release (more than 10,000), not only for data of dubious origin, but also for texts created with generative AI. In addition, unlike Frontières, nature has forbidden the use of AI to generate images and videos, precisely because of artificial intelligence software You don't cite any sources from which they derive the information with which they create the images, the verification process can therefore become very complicated.
It's possible that this exact type of dynamic was the cause of what happened with the removed article. Boundaries. For this reason, it would not be correct to view the story of the “rat with the giant penis” as a funny curiosity or an isolated mistake, but rather Symptom of a larger problem upstream. A particularly thorny problem in the scientific world, if decisive for the society in which we live. A problem that, if not solved, it could get more and more seriousAs rapid advances in artificial intelligence make it increasingly difficult to distinguish between AI-generated content and human-derived content.