Appall and disdain spread rapidly among the scientific community on Thursday as several extremely poor AI-generated visuals made rounds on social media from a peer-reviewed article recently published in a reputable journal. These visuals, acknowledged by the authors as creations of Midjourney, are completely incomprehensible. They feature nonsensical text and, notably, one includes an image of a rat with unusually large and peculiar reproductive organs, accompanied by a label reading “dck.”
The publisher of the reviewed article, Frontiers, issued an “expression of concern” on Thursday, acknowledging the raised alarms about the published content. They stated that an investigation is underway, and updates will follow upon its conclusion.
The article under scrutiny, titled “Cellular functions of spermatogonial stem cells in relation to JAK/STAT signaling pathway,” was penned by three researchers in China, with Dingjun Hao from Xi’an Honghui Hospital as the corresponding author. It was released online on Tuesday in the journal Frontiers in Cell and Developmental Biology.
Although Frontiers did not initially respond to Ars’ request for feedback, any response received will be included in this post.
The initial figure in the paper, featuring the rat image, garnered immediate attention as scientists shared and discussed it extensively on various social media platforms. Observers quickly noted the anatomical inaccuracies in the image, along with labels such as “dissilced,” “Stemm cells,” “iollotte sserotgomar,” and “dck.” Many researchers expressed astonishment and disappointment at how such a flawed AI-generated image could pass through the peer-review process and the journal’s internal checks.
However, the problematic elements extend beyond the rat image. The second figure, intended to depict the JAK-STAT signaling pathway, is equally chaotic. Instead of a clear diagram, it presents a confusing array of elements. One expert in scientific integrity even likened it to providing instructions on “how to make a donut with colorful sprinkles.” Similarly, Figure 3 offers a cluttered collection of small circular images with nonsensical annotations, aiming to visually represent the regulation of biological properties of spermatogonial stem cells by the signaling pathway from Figure 2.
Some online commentators raised doubts about whether the accompanying text was also AI-generated. Despite AI detection software suggesting its AI origin, its reliability remains questionable, as previously reported by Ars.
These flawed visuals underscore a growing issue in scientific publishing. The success of a scientist is heavily tied to their publication record, incentivizing the rapid production of articles for prestige. This environment may encourage researchers to submit low-quality work, potentially aided by AI tools. Concerns are rising that the increasing use of AI could compromise the trustworthiness of published research. Consequently, research journals have introduced new authorship guidelines for AI-generated content to combat this issue. However, as demonstrated by the Frontiers article, there are evident gaps that need addressing.
