On Thursday, some horrifying numbers generated by AI circulated from scientists’ social media networks, sparking widespread consternation and disdain. peer-reviewed paper It was recently published in a famous magazine. Although the authors acknowledge in the article that these diagrams were created by Midjourney, none of them can be interpreted. They contained gibberish text, the most striking of which included an image of a rat with grotesquely large and strange genitals, and the text “dck” his label.
On Thursday, Frontiers, the publisher of the review article, Posting an “Expression of Concern” It said it was aware of the concerns about the published work. “An investigation is currently underway and this notice will be updated accordingly once the investigation is complete,” the publisher wrote.
The paper in question is titled “Cellular functions of spermatogonial stem cells related to JAK/STAT signaling pathway” and was written by three Chinese researchers, including corresponding author Dingjun Hao of Xi’an Honghui Hospital. The paper was published online Tuesday in the journal Frontiers in Cell and Developmental Biology.
Frontiers did not immediately respond to Ars’ request for comment, but will update this post if it responds.
The paper’s first figure, the one involving the rat, quickly gained attention as scientists began sharing and commenting widely on social media platforms such as Blue Sky and the platform formerly known as Twitter. Collected. Viewed from a distance, anatomical images are clearly riddled with all sorts of mistakes. But a closer look only reveals further flaws, including labels such as “dissilced,” “Stem cell,” “iollotto sserotgomar,” and “dck.” Many researchers expressed surprise and dismay that such a blatantly bad AI-generated image could pass through their peers. – The review system and internal processes carried out by the journal.
But rat luggage isn’t the only problem. Figure 2 is less graphic, but just as broken. It’s meant to be a diagram of a complex signal transduction pathway, but it’s actually a hodgepodge. 1 scientific integrity expert He questioned whether the instructions for “how to make donuts with colorful sprinkles” were overly complicated. Like the first image, this one is full of meaningless text and cryptic images. Figure 3 is similar, a collage of small circular images filled with gibberish annotations. This image is believed to be a visual representation of how the signaling pathways in Figure 2 control the biological properties of spermatogonial stem cells.
Some scientists online wondered if the text was also generated by AI. One user pointed out that AI detection software determined that he was likely generated by an AI. However, as Ars previously reported, such software is unreliable.
Although these images are egregious examples, they highlight a growing problem in scientific publishing. A scientist’s success is highly dependent on his or her publication record, including large volumes of publications, frequent publications, and publication in prestigious journals, all of which can earn scientists greater prestige. This system incentivizes dishonest researchers to push through low-quality papers, which in the age of AI chatbots could be generated with the help of AI. Researchers are concerned that the increased use of AI will reduce the credibility of published research. Therefore, research journals recently set new author guidelines for AI-generated text to address this issue. But for now, as the Frontiers article shows, there are clearly some gaps.
