Illustration by Tag Hartman-Simkins/Futurism. Source: Getty Images
Librarians, and the books they treasure, are already fighting a losing battle for our attention with all kinds of tech-enabled brainrot.
Now, in another attack on their conscience, AI models are causing so much negligence that students and researchers keep coming to libraries and asking for journals, books and records that do not exist. scientific American reports,
one in statement From the International Committee of the Red Cross, seen by the magazine, the humanitarian organization cautioned that AI chatbots like ChatGPT, Gemini and Copilot are prone to generating fabricated archival references.
“These systems do not conduct research, verify sources, or double-check information,” the ICRC, which maintains a vast library and archives, said in a warning. “They generate new content based on statistical patterns, and therefore may generate invented catalog numbers, descriptions of documents, or even references to platforms that never existed.”
Sarah Falls, head of researcher engagement at the Library of Virginia, said: Science That AI inventions are wasting the time of librarians who are asked to search for non-existent records. He claims that fifteen percent of the email reference queries the Fall’s library receives are now ChatGPT-generated, including hallucinations involving primary source documents and published works.
“For our staff, it is very difficult to prove that a unique record does not exist,” Falls said.
Other librarians and researchers have spoken about the impacts of AI on their profession.
“This morning I spent time finding a quote for a student,” wrote A user on Bluesky who identified himself as a scholarly communications librarian. “By the time I got to the third one (with zero results), I asked where they got the lists, and the student admitted they were from Google’s AI Summary.”
“As a librarian who works with researchers,” another wrote“Can confirm that’s true.”
AI companies have focused heavily on creating powerful “reasoning” models for researchers that can do large amounts of research from a few signals. OpenAI released its agentive model in February to perform “deep research”, which it claims can do “at the level of a research analyst”. At the time, OpenAI claimed that it hallucinated at a lower rate than its other models, but admitted that it had difficulty separating “official information from rumors” and expressed uncertainty when presenting information.
The ICRC warned about that dangerous flaw in its statement. It says AI “cannot indicate that no information is present.” “Instead, they will invent accounts that appear credible but have no basis in the archival record.”
Although AI’s habit of hallucinating is well known by now, and although no one in the AI industry has made particularly impressive progress in curbing it, the technique is running rampant in academic research. Scientists and researchers, who you would expect to be as empirical and skeptical as possible, are being caught left and right submitting papers full of AI-generated citations. Ironically, the field of AI research itself is drowning in a flood of AI-written papers as some academics publish over a hundred poorly-written studies a year.
Since nothing exists in a vacuum, authentic, human-written sources and papers are now disappearing.
“Because of the large amount of carelessness that has arisen, finding records that you know about but cannot easily find without searching, has made it much more difficult to find actual records,” expressed regret at A researcher on Bluesky.
More on AI: Grok will now give directions to Tesla drivers
