Recent searches
Search options
I can totally believe this:
Dr. Damien P. William...
Remember how i told y'all that text written by neurodivergent peeps would likely have a higher false positive rate on "Al" checkers because masking means they are likely to engage in careful and specific word choice and that that would show up in text as "unnatural?" This is more of that.
Quoting Mike Masnick:
Kid has an English assignment, where school has kids first submit essay to an "Al checker." Kid did not use Al. Al checker says the use of the word "devoid" magically turned essay into 18% Al written. Changing "devoid" makes it drop to 0%. We're spending time "un-Al-ing" an essay that has no Al.
@ColinTheMathmo@mathstodon.xyz
So, if someone has a (marginally) well-developed vocabulary, their writings get flagged as AI? SOoooooo glad I'm long since beyond gradeschool and college.
THIS! While the word "devoid" isn't in my "working vocabulary" - the shrinking list of words that I routinely use to communicate - I pride myself in knowing the difference between "their", "there", "they're", and other challenges the English language presents us with. Do you suppose that correct grammar and punctuation will become AI warning flags?
@LovingFalloutLondon1954@mastodon.social @ColinTheMathmo@mathstodon.xyz
Having to define words Ive casually used in Slack or emails is a couple times a year event for me. Wife regularly uses me as a "how to pronounce" source because, as an avid reader, she sees a lot of words she's never heard anyone else use.
As a child, I was beating my college educated, avid reader parents at Scrabble starting around age 7 (and my school was encouraging me to sign up for college writing courses by about middle school).
I'd definitely have been declared a cheater if AI were scanning my essays during K-12.
@ColinTheMathmo
Literally making certain words off-limits. That's great. "Make your vocabulary 20% smaller to prove you're human."
@ColinTheMathmo race to the bottom in everything it appears
@ColinTheMathmo Recently I read an article about how to discern AI-generated spam/scam emails.
A lot of the so-called "tells" were words and phrases I use all the time, and always have.
It's going to fuck up a lot of kids to learn they sound like computers, and are therefore somehow cheating even when they're not. As an autistic kid, I remember being called on the carpet for plagarism because I used uncommon words. After awhile I started wondering if I were accidentally lying. 
@erosdiscordia @ColinTheMathmo
I feel like forcing more uniformity in human communication only makes it *easier* for AI to be passed off as original writing. How is getting humans to write like machines (uniformly, trying to sound just like everyone else) going to help us distinguish real from fake?
@artemis @ColinTheMathmo Exactly this.
Also, if people are worried about being replaced by AI, it's not the best idea to train ourselves to sound less intelligent than AI, just to distance ourselves from it.
@erosdiscordia @ColinTheMathmo It blows my mind when this happens. If you’re an adult, reading a kid’s essay, and you see an uncommon word, it’s dead easy to see if the kid is pulling it from a thesaurus because they’ve probably used it wrong. You’re an adult, you have the life experience and reading skill to recognize that. Accusing others of lying because their diction isn’t at the normal level just tells me that you aren’t reading, you’re skimming at best.
Magazines are using the same checkers against authors.
I wonder how many of my rejections, or silences without even the courtesy of a rejection, are from AI checkers.
Mostly against stories that were written before AI became a problem, and too many writers thought dictation, which I couldn’t use either, were the greatest thing since somputers.
@ColinTheMathmo or just sue the board of education.
@ColinTheMathmo Logically, there exists a set of essays written by AI, and a set of essays not written by AI. An ideal checker would determine and report which set an essay is in. So let's say we train an AI on all of the essays not written by AI, and when asked to write an essay, the AI picks an essay from that set. The ideal checker reports that the essay was not written by an Ai when it was. Hence an AI checker can not exist.
@ColinTheMathmo i have a friend who got a zero on some papers recently, apparently for using ai. She did not use ai.
@ColinTheMathmo my son is having to do this for his english class at the local community college. He periodically runs his essays through the checker as he's writing it to verify that he's not going to get fingered for using AI. Would be a whole lot easier if they just look at the edit history that comes with Word since they are required to edit and submit them with it.
@ColinTheMathmo @NilaJones I would fail this test in every possible way. I get this complaint enough from humans.
"Is this testing whether I'm a replicant or a lesbian, Mr. Deckard?"
Accusations of AI plagiarism are also linked to racist & misogynistic bias.
https://www.edweek.org/technology/black-students-are-more-likely-to-be-falsely-accused-of-using-ai-to-cheat/2024/09
https://www.washingtonpost.com/opinions/2024/02/27/google-gemini-bias-race-politics/
https://www.washingtonpost.com/technology/2024/02/22/google-gemini-ai-image-generation-pause/
@ColinTheMathmo I got ridiculed in middle school for knowing the word “ridicule”, and got roasted by my whole class in elementary school because I wrote a two page report about FDR because I had read a book about him and wanted to share all this knowledge I’d gained. Took me years to get back to writing. I’ve never used AI anything.
It fucking enrages me when demonstrable skill is deemed suspiciously uncommon. Fuck you for trying, right? Get back in line.
@ColinTheMathmo ahh that stinks, I love the word devoid. I also love ‘delve’ cant use that one now either.
so my verbosity with expansive lexicon that seeks to deploy fun/esoteric/anachronistic words instead will raise a flag.
And peppering it with swears like how I talk out loud doesn't help.
Fascinating stuff.
@ColinTheMathmo Ai is getting stupider as the human beings creating or trying to use it get madder. If only the term 'intelligence" hadn't been used. #AI #generativeAI
@ColinTheMathmo I.e. anything written by non-native speakers will likely be labelled as AI, because we tend to use words that sound similar to words in our native languages, which are often perfectly correct, but also rarely used words.
@ColinTheMathmo interesting. Academia tends to reward using overly complex language but maybe this will change due to LLMs. Then they may have no choice but to actually look at the ideas being conveyed.
@ColinTheMathmo If you're ND or otherwise at risk of being called AI and smeared, do yourself a favor and use OBS to record your screen as you write. Then if the AI hammer comes down, file a lawsuit with your proof in hand.
@Lydie This. Audit trails are becoming critical.
Whilst I understand the sentiment.. my life experience makes me acutely aware of the onus of proof being on the victim of bias yet again.