Hacker News Reader

Comment by userbinator | Hacker News Reader

"eneeavor" "kalsomining" "dodr" "Decline I."

Is this AI-generated? It certainly has all the signs of being so.

I've read real books from the late 19th and early 20th century, and while occasional typos do appear, their density here is suspicious.

Thus my conclusion is that I don't think this is a real 8th grade test.

View on HN

Real or fake, it's not AI generated. It's been around for over a decade (this same one, with those typos).

https://www.bullittcountyhistory.com/bchistory/schoolexam191... - on the answers page they also acknowledge the typo on "eneeavor".

https://headsup.scoutlife.org/would-you-pass-this-test/ - 2013

Thank you for finding the truth. Books from that era would likely have been subjected to far more careful proofreading, which explains the typo density (or lack thereof) compared to this.

“Decline I” is an instruction for the student to provide the first person pronoun in all cases: I (nominative), me (accusative/dative/ablative), my (genitive), mine (genitive substantive). (I have borrowed the case names from Latin, with which I am more familiar. I think the English cases are nominative, objective, possessive.)

I believe the misspellings in the spelling section are intentional so that the student will identify them—I am guessing that’s the point.

"eneeavor" is I think the only misspelling in the spelling section, so I don't think it is intentional. Perhaps the test was read aloud.

"Kalsomining" is the -ing form of "kalsomine", ie. whitewash.

"Decline I" is a request to decline the pronoun "I".

I'll accept these as being legitimate, but that doesn't explain "dodr" nor "eneeavor".

Edit: there's also "secrate" and "Pres dent"

They do appear to be simple typos.

edit: The test seems to come from this page from the county's museum: https://www.bullittcountyhistory.com/bchistory/schoolexam191.... The WayBack machine records the first snapshot in 2012, so AI seems unlikely.

They note the spelling errors

> Note that there are several typesetting mistakes on the test including a mistake in the spelling list. The word "eneeavor" should be "endeavor." This version of the exam was probably a master version given out to the schools (note that the spelling words wouldn’t be written on a test.) The museum has been told that the exam was handed out in a scroll form (that is why the paper is long.) The typos would have been corrected simply by contacting the teachers and telling them to mark their copies accordingly, much like would be done today. And there might not be quite as many typos as you think; "Serbia" for example was indeed spelled "Servia" back then.

>Is this AI-generated? It certainly has all the signs of being so.

Current generation LLMs does not make smelling mistakes, only humans do.

At least in Finnish I’ve seen some, not often but they are there.

I think that the person was talking about image generation models.

Even the best of existing image generation models often spell words hilariously wrong when asked to generate an image containing text.

Correct, they ddon't'

But sometimes their messassage gets mangled in intersting ways

What is "Decline I" a typo of? I interpret is as a request to decline (that is, give the inflected forms of) the pronoun "I".

Could also be "Decline one", in any case, not a typo. "I" can be declined as subjective I, objective me, reflexive myself, possessive my, independent possessive mine.

I've never seen an LLM typo a word in normal usage. Who knows if it's fake, but I highly doubt it's AI-generated.

Generative AI tends to produce this sort of output.

https://cdn.arstechnica.net/wp-content/uploads/2024/02/fcell...

It's true that LLMs are very good at reproducing correctly spelled text - but we're looking at an image here, and image-generating AIs aren't so good at generating text in most cases. They've gotten a lot better recently, but they're still not quite there yet.

(To be clear, I don't believe this is actually generated by AI, but in this case, but since we're looking at an image, I can see why some people might be concerned that it might be AI given the misspellings.)

If it were a generated image there would be giveaway malformed/deformed letters, but these letterforms are immaculate.

But yes in general it's a reasonable concern to have.

Reply on news.ycombinator.com