23,000 dissertations were filtered through Pangram, which has a 1-in-10,000 false positive rate. It turns out that ONE IN FOUR dissertations contain AI prose kek.
>>25310897Odds that 1 out of 4 dissertations were "written" by indians?
>>25310903Indians tend to have better English than Chinese and both groups are notorious for cheating, so I suspect it’s a mix
>>25310897What percentage of dissertations in the 20th century were written using this methodologyEven the pope got caught using AIhttps://www.theverge.com/ai-artificial-intelligence/937801/pope-leo-xiv-magnifica-humanitas-ai-pangram>It’s possible that AI was used to write parts of Pope Leo XIV’s latest encyclical about AI’s impact on humanity. An analysis by Linch Zhang posted on the forum LessWrong found certain paragraphs of Magnifica Humanitas to be between 40 percent and 100 percent written by AI, according to the popular AI detector Pangram.>The document includes known traits that appear in AI-generated writing, such as a higher use of the word “genuinely” — which crops up in writing by Anthropic’s Claude — than previous encyclicals, Zhang says. Another person ran the text of the document section by section through Pangram, finding that 62 percent of its first chapter was flagged as AI generated. When The Verge ran roughly 2,000 words of the document through Pangram, it estimated that 46 percent was AI-written.
>>25310903If you write retarditly you are less likely to trigger AI detection software it's when you write gay that you get called an AI
>>25310897>Pangram, which has a 1-in-10,000 false positive rate.Bullshit. I just tested it on two of my own texts, it couldn't detect the AI assisted one or the entirely AI+light editing one. I have no doubt AI dissertations are flooding the world, especially counting anyone smart enough to work the material. This program is as useful as flipping a fucking coin.
>>25310958>What percentage of dissertations in the 20th century were written using this methodologyClose to zero. Virtually no pre-2022 documents come up as AI-written on Pangram.
>>25310897basically llms are only for background knowledge and theyre useless when you start asking expert level questions, but, 100% of academics are subhuman garbage
>>25311006>t. high school dropout chud
I'm pretty sure someone in my writing group is using AI >she joins the group and right away it's clear she's autistic >asks us all to be brutal because she's so tired of people telling her what a wonderful writer she is >also tells out flat out with a huge smile on her face about what a dark traumatic past she has in real life about getting kidnapped and raped. So I know for a fact that she's a habitual liar. >when it's her turn to critique she either says she didn't read our stuff or she gives obviously fake critique that makes it clear she didn't read it>read her stuff and she has no idea how to tell a story >doesn't know how to describe stuff>every character and location has nearly identical names like Xardar, Xamdam, Xarwrdoor, Xaaraawr...>every character is introduced by the main character thinking about them, not by them actually being in the story or doing something. For example, she'll just think about her sister and how she misses her, and that's how we meet her sister seconds before she appears and then behaves in a completely out of character way.>just expects you to understand how magic and leprechauns work with zero explanation>rawdogs the reader with a bunch of pointless world building infodumps about forever wars and ongoing genocides >she doesn't write in paragraphs. Everything is a super dramatic single sentence that has its own line...>Like this. >a month later she submits some rewrites >it has none of these problems anymore and has the rhythm of AI prose >I don't think it's 100% AI but it's very clear that she's using it to punch up descriptions, feed her metaphors and similes, and add dramatic weight to scenes. maybe I should check her submissions for hidden AI artifacts.
>>25311004>Virtually no pre-2022 documents come up as AI-written on PangramHow many have you personally tried?
>>25310969>couldn't detect the AI assisted one or the entirely AI+light editing one.That's called a false negative not a false positive, illiterate anon fren
>>25311044>just expects you to understand how magic and leprechauns work with zero explanationit's magic you ain't gotta explain shit
>>25311068I'm not saying a magic system is necessary. I'm saying she expects us to know what "the leprechaun fire" is and what it did to the main character.
>>25310958It literally flagged interview quotes in my work as AI. The detectors are as garbage as the AI text itself. Any talk of "muh benchmarks, muh rates," is currently unregulated and should be treated like gas station supplement market copy.
>>25311004I've run my old papers from before AI through and they get sections flagged, sometimes for very odd bits.
I just tried Panagram and interestingly it flagged some sentences as likely AI, but then when I did the same text as part of a larger excerpt it decided all of it was likely human. I highly doubt their claimed statistics are accurate considering I got it to foul up on a single try.
>>25311128I managed to get a second false positive simply by replacing commas with em-dashes for a clause.Damn, who woulda thunk an AI company would just go out there and tell lies on the Internet about the effectiveness of their product.
>>2531100412.5% of the opening paragraphs of Fides et Ratio came up as AI on ZeroGPT. I'm out of credits for the new one. These generally suck.
>>25310897>>25310958The most worrying part about all this for me is when people's everyday writing style -- subconsciously or otherwise -- starts to genuinely mimic AI prose.But yeah nah, jokes aside because just down wind of that (how people write) is how they think.
>>25311150It's already happening and through AI edits, people are blindly accepting AIs suggestions without thinking things through
>>25310958The pope doesn't actually write the encyclical; apparently his live speeches came up human written, so it's probably not him or his speech writers who used AI. Encyclicals are a collaborative project with many different people: bishops, specialists, etc. IMO, most likely some individual used AI without anyone else's knowledge. I think someone pointed out that some of the pope's tweets came up AI, but I doubt he actually uses X.
>>25311389When Pope Francis was alive, the official account on a social media (Can't remember if it was X/Twitter or something else) liked some ethot's post. Normally I'd think, "oh they put some intern in charge of it," but this is the Vatican. Surely they would have had a priest or monk in charge of it.I wonder if anyone else at the Vatican noticed it lol
All these AI "detectors" are dog shit. You can easily get it to recognize 100% human text as AI, and vice versa. The AI detector is funnily enough an AI jeet scam in and of itself.
>>25310897>seem to be
It's mathematically proven that it's impossible to see if a text is AI generated or not.
>>25311044>he doesn't know how leprechauns work
>>25311451> The AI detector is funnily enough an AI jeet scam in and of itself.This is correct and the tech illiterates in the anti AI camp just out themselves promoting another AI product. It’s quite insane but illustrates why AI will win. The average retard can’t understand what it is. Too many people blindly accept the slop even at current levels. Even looking at the local notice boards it’s slop, the average idiot wants it, uses it, can’t tell the difference. It’s like talking about how motion smoothing is making their TV content look like shit and they just stare blankly as you turn it off and on. Or people with high refresh rate monitors who never actually put it in the high refresh rate and swear it’s better.
>>25310897THATS IT!Every single paper needs to be written by hand or it must be assumed it is AI.
>>25311150in this case it would be AI making the poorly-educated rabble familiar with the role of em dashes in grammar. em dashes were always used extensively among the educated class, and AI trained on published material, so it quickly learned to use em dashes. but poor, unededucated zoomers who only knew "english" from twitter had never seen this thing before and saw it as an artifact of computer-generated thought. and here we are now, with even the new york times putting in spaces between the dashes, because new york times writers are otherwise going to be accused of using chatgpt apparently.
>>25311887I'm not talking about the dashes dipshit
>>25310929SAAAAR>>25311578Link to the article?Considering that LLM source mostly from reddit and that reddit users have a very particular writing style, I doubt it.
>>25311578it's also been mathematically proven that you're a flaming homosexual
>>25311044>just expects you to understand how magic and leprechauns work with zero explanationany explanation will be worse than what your mind would have filled in, recently read a furry macro weight gain and transformation fetish story that had a lengthy digression explaining dragons are immune to the square cube law due to the relativistic structure of their DNA that might as well just not have been written.
