A review article with some obviously fake and non-scientific illustrations created by Artificial Intelligence (AI) was the talk on X (Twitter) today.
The figures in the paper were generated by the AI tool Midjourney, which generated some pretty, but nonsensical, illustrations with unreadable text.
It appears that neither the editor nor the two peer reviewers looked at the figures at all. The paper was peer-reviewed within a couple of weeks and published two days ago.
Dear readers, today I present you: the rat with the enormous family jewels and the diƨlocttal stem ells.
The paper by Xinyu Guo et al., Cellular functions of spermatogonial stem cells in relation to JAK/STAT signaling pathway, Frontiers in Cell and Developmental Biology 2024, DOI 10.3389/fcell.2023.1339390 [link to PDF, in case the publisher removes it], easily passed editorial and peer review.
The authors disclose that the figures were generated by Midjourney, but the images are – ahem – anatomically and scientifically incorrect.
Figure 1 features an illustration of a rat, sitting up like a squirrel, with four enormous testicles and a giant … penis? The figure includes indecipherable labels like ‘testtomcels‘, ‘senctolic‘, ‘dissilced‘, ‘iollotte sserotgomar‘ and ‘diƨlocttal stem ells’. At least the word ‘rat‘ is correct.
One of the insets shows a ‘retat‘, with some ‘sterrn cells‘ in a Petri dish with a serving spoon. Enjoy!
Figure 2 appears to show an impressive scientific diagram of the JAK-STAT signaling pathway. Or does it explain how to make a donut with colorful sprinkles? Again the words and numbers are made up. What do ‘signal bıidimg the recetein‘, ‘Sinkecler‘, ‘dimimeriom eme‘, ‘Tramioncatiion of 2xℇpens‘, ‘ↄ‘, and ‘proprounization‘ mean? [my spell checker is getting very angry with me].
Figure 3 appears to show a bunch of pizzas with pink salami and blue tomatoes.
Of course, we can have a good laugh at these figures, and wonder how on earth the handling editor and the two peer reviewers didn’t catch this.
But the paper is actually a sad example of how scientific journals, editors, and peer reviewers can be naive – or possibly even in the loop – in terms of accepting and publishing AI-generated crap. These figures are clearly not scientifically correct, but if such botched illustrations can pass peer review so easily, more realistic-looking AI-generated figures have likely already infiltrated the scientific literature. Generative AI will do serious harm to the quality, trustworthiness, and value of scientific papers.
The Tadpole Paper Mill papers – a set of 600 fabricated papers from the same design studio – were perhaps one of the earliest examples of peer-reviewed papers containing computer-generated images of Western blots. We were able to identify them as fakes because all blots had the same background.
But recent advances in AI technology mean we’re already past the stage where a human can distinguish a fake photo from a real photo. Just take this recent New York Times quiz to see if you can spot the difference.
I don’t know, I thought Figure 3E gave a great summary of immouminomuduodiuilatiuciaton!
LikeLike
I’m glad that the increased exposure might end up attracting a whole new generation of young iollotte sserotgomar biologists, though.
LikeLike
The pictures are hilariously awful, but the text smells of LLM, too. For example, there’s a sudden swerve into what sounds like a specific experiment, which doesn’t fit with the text before or after and is very strange in a review: “For the cell culture aspect of SSCs, there are quite a few experiences to learn. SSCs were isolated from the testes of C57BL/6J mice or SD rats on the 7th day after birth. Subsequently, after removal of the tunica albuginea and epididymal curvature,…”
And then we get: “They were physically sheared and digested with a solution of DnaseI, hyaluronidase, collagenase, and trypsin using a two-step enzymatic digestion method in which the digestive enzymes included DnaseI, hyaluronidase, collagenase, and trypsin.”
LikeLike
Published in “Frontiers in Cell and Developmental Biology”? Somehow I am not surprised.
LikeLike
I ran portions of this in a few AI detectors, and Quillbot and Copyleaks found it entirely 100% AI generated.
LikeLike
From my point of view, scientists are 10% genius, 90% hard work, although many of them think their percentage of genius (and ego) is close to 100. In this case, it is confirmed that anyone can be a fool regardless of academic degrees. Common sense is something that most of humanity suffers from. It would be good to locate the clowns who passed such a barbarity, just out of morbid curiosity. Anyway, it is incredible how visual communication is underestimated: For example, the lamentable “graphical abstracts” that appear in many papers, it seems that the authors present them without interest and by mere requirement, totally ignorant of the potential of an image.
LikeLike
Author affiliations…dept of spine surgery. Why on earth would they be publishing in this field anyway? Thank goodness it got flagged and there are some utter crap AI indicators that can pull some work. Scary that this got through, so others that have been slightly improved stand an even better chance.
LikeLike
I’ve just got published a 3-years research on the AJSM.
The thoroughness and “human” attention to details of the editing process made me reflect on one possible way out to avoid examples like the rat above: pursuing quality, not quantity.
It means trying to work hard to reach the journals which provides a rigorous review process, the only way to avoid cases like this.
LikeLike