A Scanning Error Created a Fake Science Term—Now AI Won’t Let It Die

An AI error has caused nonsensical phrases like "vegetative electron microscopy" to appear in scientific papers and AI responses, originating from a digitization mistake in a 1959 paper. This issue is exacerbated by AI models replicating the error. Researchers have identified the CommonCrawl dataset as a likely source of the problem but removing such errors is challenging due to the vast amount of data involved and lack of accessibility to training datasets from major companies. Publishing giants like Elsevier initially defended the term's validity before issuing corrections, highlighting the broader issue of "junk science" in scholarly databases. This underscores the challenges AI faces in maintaining accurate scientific information at scale.