A brand new type of glitch: Errors that only AI can introduce…
Shakespeare tells us that Anthony’s funeral oration at Julius Caesar’s funeral featured the words, “The evil that men do lives after them; The good is oft interrèd with their bones.” In our context today, we forget how things that work originated; we are all too often reminded of what didn’t work.
At RealClearScience, Aaron J. Snoswell, Kevin Witzenberger, and Rayane El Masri tell us about a weird phrase that means nothing but now infests science papers, thanks to AI:
Earlier this year, scientists discovered a peculiar term appearing in published papers: “vegetative electron microscopy”. This phrase, which sounds technical but is actually nonsense, has become a “digital fossil” – an error preserved and reinforced in artificial intelligence (AI) systems that is nearly impossible to remove from our knowledge repositories.
“A Weird Phrase Is Plaguing Science Papers Thanks to AI,” April 16, 2025
How did it get there? Through “ a remarkable coincidence of unrelated errors.”
First, two papers from the 1950s, published in the journal Bacteriological Reviews, were scanned and digitised.
However, the digitising process erroneously combined “vegetative” from one column of text with “electron” from another. As a result, the phantom term was created. “Is Plaguing Science Papers”
Then, due to a a translation error, the phrase turned up in some Iranian papers. It ended up in 22 papers, one of which was retracted and one corrected, and it appeared in investigations about science publishing integrity. The error apparently persists in large language models, which, say the authors, “suggests the nonsense term may now be permanently embedded in AI knowledge bases.”
Digital Fossil
At Gizmodo, Isaac Schultz calls it,
… a textbook case of what researchers call a digital fossil: An error that gets preserved in the layers of AI training data and pops up unexpectedly in future outputs. The digital fossils are “nearly impossible to remove from our knowledge repositories,” according to a team of AI researchers who traced the curious case of “vegetative electron microscopy,” as noted in The Conversation.
“A Scanning Error Created a Fake Science Term—Now AI Won’t Let It Die,” April 178, 2025
It’s a good early warning about the risks of letting too much depend on AI. We can be glad it didn’t happen at the International Space Station or in a self-driving car where the output might be critical.