[ad_1]
You might think of this as a massive AI cock-up.
A scientific paper purporting to demonstrate a sperm stem cell signaling pathway has been met with widespread ridicule after depicting a rodent with an anatomically striking appendage and four giant testicles.
The creature, labeled “Rat,” also sat upright like a squirrel, but its graphics included gibberish words such as “dismantled,” “testtomsel,” and “centric.” It was scattered.
The cropped image showed “stem cells” in a Petri dish being lifted out with a spoon.
The image, along with several other absurd graphics generated by the AI tool Midjourney, was published this week in the journal Frontiers in Cell and Development Biology.
It includes a colorful diagram of the JAK-STAT signaling pathway, which experts liken to “some crazy level of Candy Crush” and say is not based on “known biology.” Ta.
The paper, written by researchers at China’s Honghui Hospital, was later retracted by the journal, which apologized and said it was working to “correct the record.”
The dangers of fabricated research
But some scientists have raised concerns about how the paper was published in the first place, warning that it ushered in a dangerous era in which researchers use AI to falsify research. There is.
Adrian Liston, professor of pathology at the University of Cambridge and editor of the journal Immunology & Cell Biology, said: It does not check whether they are correct.
“It’s like actors who play doctors on TV shows. They look like doctors, sound like doctors, and even use the words that doctors use. But you don’t get medical advice from actors. You don’t want to.
“In my opinion, the JAK numbers are even worse than the rat numbers. There are just meaningless connections and lines that have nothing to do with known biology.”
He added: “The real journal problem is becoming increasingly difficult as generative AI makes fraud easier.”
“It used to be very obvious to tell a cheat sheet at a glance. It’s becoming increasingly difficult, and many people in scientific publishing are at a point where they can’t manually tell whether a paper is real or a scam. I’m really concerned that we’re going to reach a point.”
Other scientists on social media described the graphic as “absolutely shameful” and “devastating”, with one saying: “I don’t know whether to laugh or cry.”
Professor John Tregoning of Imperial College London said the graphic was “objectively interesting” but “has no place in a scientific journal”.
Laughter
Writing in Science Integrity Digest, Dr. Elisabeth Bick, a Dutch microbiologist who works on discovering manipulation in scientific papers, says: The reviewers couldn’t figure this out.
“But this paper is actually a sad example of how naive scientific journals, editors, and reviewers can be in accepting and publishing AI-generated crap.
“These numbers are clearly not scientifically correct, but if a botched illustration like this can easily pass peer review, it is likely that more realistic-looking numbers generated by AI are already making their way into the scientific literature. There may be.”
Dr. Bick has identified more than 1,000 papers containing fraudulent images, most of which she believes were generated by AI.
Companies are trying to develop software to detect AI-generated content, and it may be watermarked in the future, but for now, there’s no way to tell unless it’s clearly wrong. there is no.
A spokesperson for Frontiers in Cell and Development Biology said: “We would like to thank our readers for reviewing our paper. If we are wrong, the crowdsourcing dynamics of open science allow us to provide feedback from the community. will help you correct your records quickly.”
The Telegraph has contacted the authors for comment.
[ad_2]
Source link
