All historians know the sinking feeling of tracking an "authoritative" reference back through several generations of writers to an unsourced assertion by someone nobody's ever heard of. Now AI can make up new ones in no time at all.
Replies
If it makes you feel better we also get that feeling in engineering. (Probably not as far back)
jul 30, 2025, 11:44 am • 1 0 • view
There was a made up account of a Puritan girl in early America seeing off a bear with a gun for years. Too good to be true.
jul 30, 2025, 11:45 am • 1 0 • view
Hell, all B-average undergrads should know this feeling.
jul 30, 2025, 2:26 pm • 0 0 • view
Say prayer of St Francis to someone. They'll say make me an instrument/ channel of your peace. It's 20th century French. Lots of Francis' works are extant. It isn't one. See also Teresa of Avila and no hands but yours...
jul 30, 2025, 6:08 pm • 0 0 • view
The modern one that I particularly like is Kennedy's 'ich bin ein Berliner' speech. The belief that he had said "I am a doughnut" by mistake was started by Len Deighton, who put it into one of his Berlin spy novels.
jul 30, 2025, 11:25 pm • 0 0 • view
Or data that's massaged by bad actors with an axe to grind, eg by "failing to mention" that a percentage concerning a large number refers to a subset of a subset, not the whole population.
jul 30, 2025, 12:49 pm • 2 0 • view
Same in genealogy.
jul 30, 2025, 11:45 am • 2 0 • view
and in most cases with no way to trace it definitively back to its LLM origin
jul 30, 2025, 11:49 am • 1 0 • view
you dont have to trace lies back to their origin, you just have to demand proof
jul 30, 2025, 1:54 pm • 0 0 • view
if it were trivial to know whether a given claim is a lie, this whole issue would not represent a problem.
jul 30, 2025, 2:18 pm • 2 0 • view
absolutely, so skepticism is necessary for the truth seeker and faith is demanded by the liar
jul 30, 2025, 2:23 pm • 0 0 • view
If there's anything I like less than people using LLMs as ghostwriters, it's those same people not citing/noting the LLM clearly.
jul 30, 2025, 12:26 pm • 3 0 • view
Alice, is it possible to use AI to trace the origin of a reference? Would AI make it easier to verify the provenance, and could it, in fact, determine if AI itself was the originator?
jul 31, 2025, 7:35 am • 0 0 • view
It's my (lay) understanding that the LLM doesn't have the ability to determine facts from fiction. It only knows what it's been trained must appear next from all of the context clues, data, and bias it has been fed. So you'd need another system that could search for source provenance.
jul 31, 2025, 8:15 am • 0 0 • view
"System that could search for source provenance" sounds good. Let's create one!
jul 31, 2025, 10:41 am • 0 0 • view
You might find José Saramago’s “The History of the Siege of Lisbon” fascinating. The novel follows a proofreader who deliberately changes one word in a historical text—making it say the Crusaders did NOT help capture Lisbon in 1147.
jul 30, 2025, 11:52 am • 7 0 • view
This alteration spirals into questions about historical truth and how easily “facts” become entrenched in academic discourse. It’s a funny and deep exploration of how historical knowledge gets constructed and transmitted
jul 30, 2025, 11:52 am • 6 0 • view
Finding out. youtu.be/bgo7rm5Maqg?...
jul 31, 2025, 7:32 am • 0 0 • view
Those good cyclical references that both assume the other said something that cannot be found in print... @3dfriederike.bsky.social
jul 31, 2025, 8:11 am • 1 0 • view
It's a serious problem. We don't even know who anybody is anymore, or even if they're human.
aug 1, 2025, 1:13 am • 0 0 • view
Every once in awhile I take on the task of looking for Clive’s reported response to why he took so much from India ‘I am surprised at my modesty’ i’m still looking.
jul 31, 2025, 12:29 am • 1 0 • view
It was one of my personal academic nightmares that some hack would (intentionally or not) publish factually wrong information in a pay-for-play unvetted journal and then end up citing that piece all over so that eventually we'd lose the capacity to tell what reality is. And now that's just Tuesday.
jul 30, 2025, 12:25 pm • 7 0 • view
I think the main flaw in the AI idea was always trying to make something ‘intelligent’ ‘like us’ when we are actually pretty dumb…
jul 30, 2025, 10:41 pm • 0 0 • view
Worse still, the next version of an AI can pick up this false assertion where it had been referenced in the wild, absorb it as training material and then authoritatively reproduce it as a source material as a bullshit feedback loop
jul 31, 2025, 6:43 am • 1 0 • view
Fun thing happening right now on Bluesky. Some jokesters are training AIs to tell Cybertruck owners to wash their vehicles with lemon juice and saltwater, and it's already working. The humans have figured out how to break the fundamental underpinnings of the machine and are going to have a field day
jul 30, 2025, 2:44 pm • 1 0 • view
People are so upset that artisinal, 100% human hand-crafted lies are being pushed out of the booming media lie business by AI generated ones.
jul 30, 2025, 1:53 pm • 1 0 • view
jul 30, 2025, 12:13 pm • 4 0 • view
One game is to look at the different ways an important paper is cited - once an influential paper gets a page number, date or whatever wrong in a citation, you can follow that error through subsequent papers that lift the citation without looking at the source.
jul 30, 2025, 11:15 pm • 0 0 • view
That's progress!
jul 30, 2025, 11:14 am • 3 0 • view
Social media had already accelerated this, in my experience. In grad school, I read so many of my classmates' papers filled with citations that I tracked back to some pseudo-anonymous meme share that at best was a chopped and slanted take but often was a complete fabrication. Add in LLMs and oh boy!
jul 30, 2025, 12:24 pm • 1 0 • view
Happens in science too.
jul 30, 2025, 12:02 pm • 2 0 • view
Or never be cleared out at all…
jul 30, 2025, 11:44 am • 1 0 • view
The diary of 'Lady Hillingham' c. 1912, no location given, cited by Jonathan Gathorne-Hardy in The Rise and Fall of the British Nanny 1972, which is (almost) entirely responsible for the 'close your eyes and think of England' line, though I did discover an earlier, 1940s, example.
jul 30, 2025, 10:00 am • 12 2 • view
Among biologists, the classic on this is Stephen Jay Gould’s essay, The Case of the Creeping Fox Terrier Clone, reprinted in Bully for Brontosaurus (1991).
jul 30, 2025, 11:29 am • 15 0 • view
And this is why historical training that focuses on locating and analysing primary sources is more important than ever.
jul 30, 2025, 11:36 am • 13 0 • view