In the vast and ever-expanding world of scientific publishing, a bizarre phrase has infiltrated dozens of academic articles—and artificial intelligence may be the culprit. The term “vegetative electron microscopy” has surfaced in at least 22 peer-reviewed papers, despite having no technical or scientific meaning whatsoever.
A cascade of errors from OCR to translation
The origins of this nonsensical phrase trace back to a 1959 article where “vegetative” and “electron microscopy” appeared in separate columns. Optical character recognition software (OCR), widely used to digitize old texts, appears to have mistakenly combined them. Another plausible cause lies in Persian-language translation, where “scanning” and “vegetative” differ by a single dot in writing, leading to systematic mistranslations.
AI models perpetuate the nonsense
Once the error entered the digital sphere, it didn’t stop there. Language models like GPT-3 and Claude 3.5 trained on contaminated data absorbed the phrase as if it were legitimate. These models, unable to distinguish credible from flawed input without human oversight, now reproduce the phrase confidently in scientific-style outputs, compounding the issue.
Editors and reviewers failed to intervene
Perhaps the most troubling aspect is that major academic publishers like Elsevier and Springer Nature validated these articles, either through negligence or indifference. In some cases, reviewers defended the phrase with implausible justifications, further embedding the error into the scientific canon.
A symptom of a deeper crisis
This isn’t an isolated blunder. It highlights a wider crisis in scientific publishing, where paper mills, poor translation, and unchecked AI usage are eroding the integrity of research. The question remains: how many more errors like this are silently shaping our understanding of science?