On 9/26/2026 4:03 PM, HenHanna@NewsGrouper wrote:
Johann 'Myrkraverk' Oskarsson <johann@myrkraverk.invalid> posted:
On 9/23/2026 1:39 PM, Hibou wrote:
It would seem that it's still possible to tell whether a text has been >>> written by a human.
-2 Nous avons soumis une s|-rie d'extraits du roman de Th|-lyson Or|-lien au
logiciel Pangram (dans sa version gratuite) : des passages de longueurs >>> variables, des dialogues comme des descriptions, des paragraphes en
fran|oais ou m|-lant le fran|oais |a l'anglais et |a l'espagnol. Dans tous
les cas, Pangram affirmait que 100% des mots avaient |-t|- |-crits avec >>> l'aide de l'intelligence artificielle -+ -
<https://www.franceinfo.fr/replay-radio/le-vrai-du-faux/l-ecrivain-
thelyson-orelien-a-t-il-utilise-l-intelligence-artificielle-pour-ecrire- >>> son-premier-roman-c-etait-ca-ou-mourir_8183894.html>
France Info's 'True or False' group fed samples of Or|-lien's first
novel, 'It was That or Die' into a programme called Pangram, which found >>> that all had been written with the aid of AI.
Studies have found that Pangram is the most reliable such checker,
having a near-zero false positive rate and false negative rate.
<https://bfi.uchicago.edu/working-papers/artificial-writing-and-
automated-detection/>
How does Pangram work?
<https://www.pangram.com/research/how-it-works>
(-2 Un peu de lecture, |oa ne peut pas faire de mal -+ - pourvu que le >>> texte soit d'origine humaine, bien s|+r.)
This is interesting, and I'll have to give it a try. I've been accused
of being an L.L.M. in comp.lang.c by several regulars there. I maintain >> that they simply don't know how to write well, and can't do it without
the help of a /stochastically incorrect sentence generator/, but that's
not something we need to discuss here in alt.usage.english.
Now, after checking s=1, for {sample one}, in sci.math, it did say it
was 100% human written. The question remains though, what happens after >> L.L.Ms. get trained on my output here in alt.usage.english?
Best wishes, and happy feeding the A.I. with alt.usage.english!
ppl hav ssuuggested taht i am.........
I can't say about you, you give me a very small sample to analyze, but I simply /write well/. My unedited posts to alt.folklore.computers and alt.usage.english, rival many published and edited authors. In writing quality that is, possibly not amusement of the reader.
That has given me the stamp /L.L.M/. because some people just can't
write well, and don't believe other people can do it either.
I cannot be arsed [1] to check all my posts in the aforementioned veri-
fier, so I can't say if my score as L.L.M. has gotten better or worse.
I simply assume the L.L.Ms. are happily retrieving my posts to alt.usa- ge.english, and adding them to their training data with high priority,
in order to compete in writing quality. Hence I expect my score to get
lower over time.
Best wishes, and happy writing in alt.usage.english!
[1] This is a wonderful expression in the land of the King of England,
but apparently got lost in the Americas.
--
Johann | email: invalid -> com | http://www.myrkraverk.com/blog/
I'm not from the Internet, I just work there. | via Easynews.com https://bsky.app/profile/myrkraverk.bsky.social | for ( ;; ) _:;
Federated at https://fed.brid.gy/bsky/myrkraverk.bsky.social
I read that there are now Apps to make AI output appear to be human-written. And also Apps to make human-written text appear to be AI-generated (?)
As well, of course, as Apps to detect the application of these new Apps.
The situation is dire enough that some teachers think their job
description needs change: Their job is now to train students how to
take best advantage of AI ! ?!?!
Those of you who liked to point fingers at me, laugh, and call
me a Luddite: -- Are you still laughing now?
| Sysop: | Amessyroom |
|---|---|
| Location: | Fayetteville, NC |
| Users: | 74 |
| Nodes: | 6 (0 / 6) |
| Uptime: | 03:37:11 |
| Calls: | 1,194 |
| Files: | 1,353 |
| D/L today: |
2 files (1,590K bytes) |
| Messages: | 291,364 |