[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: dissertation.png (64 KB, 722x587)
64 KB PNG
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.
>>
File: 1777036836824912.jpg (10 KB, 319x319)
10 KB JPG
They're probably better than the human ones. Comfy slurp.
>>
>>108937844
Kwab
>>
>>108937844
>Pangram, which has a 1-in-10,000 false positive rate
yeah right
>>
>>108937884
Virtually no texts written prior to November 2022 are AI-flagged.
>>
So what’s the problem, really? Are the dissertations wrong?
>>
>>108937905
no problem
in fact they should start giving GPT PhDs, and let zoomers die in a ditch
>>
>>108937905
>It's okay if you pay someone to write your dissertation, as long as it's not wrong.
>>
>>108937844
You'd be retarded to not use AI for this. Using AI doesn't mean there has to be no work or thought involved.
>>
>>108937930
You'd be retardedto not hire someone to write your dissertation. Hiring someone to write your dissertation doesn't mean there has to be no work or thought involved.
>>
>>108937928
It’s a tool. It’s not a someone.
>>
>>108937978
You are a retard.
>>
>>108937901
Which is meaningless if it uses the date of publishing as a heuristic.
>>
>>108937928
you can continue using gippity in your professional life so it's fine.
>>
>>108937935
You mean work together* to write your dissertation. There's a difference between getting someone else to do all the work and only being able to take the final copy, versus working with someone to see how your work should be written, proof read it, or even write out the things you tell it to. Generative AI is a powerful simplification tool for most things in real life, but it is also a "more in, more out" amplifier. The more you know yourself and the more you direct it, the better results you get.

In example,
>Write a paper about cats.
is going to be infinitely worse than
>Write a paper about about cats, starting with the history of their domestication, move into pest control work before modern extermination, then household benefits on happiness, and finish on risks of toxins from their feces on both pregnant women and non-pregnant owners.
One is structured to your intent, the other is going to be who knows what. And assuming someone will put in any of the slightest work, they'll be nudging direction as it goes and asking it to incorporate a few gathered studies.

And that's still just the minimal human effort. More in can be bullet point outlines getting flushed out, or rough drafts being corrected, or asking to strengthen only certain portions of a paper. It could also have no AI in the text of the paper, only summaries of data tables or different perspectives on conclusions from data before you start writing. How much AI is used versus the human can vary, despite all being AI-written in some capacity. These have the same inverse in human aid as having a tutor or partner to work with, and also whole writers. It's a spectrum.
>>
>>108937844
wanna bet how many are written by a Zhang or a Singha
>>
>>108937844
>which has a 1-in-10,000 false positive rate

Press X to doubt
>>
>>108937884
This. Ppl see so much AI prose these days, both knowingly and not, it's bound to influence their own writing style eventually, some faster than others. I have zero doubts there are students out there rn who've already picked up a few AI prose mannerisms.
>>
>>108937844
Aigods won
>>
>>108937844
meh, people used to cheat by paying someone else instead of having a machine do stuff for them. e.g.: some years ago, I helped my boss write some shit for some postgrad course in exchange for some cash.
at least you can detect cheaters these days.
>>
>>108937844
>should I start a witch hunt because im bored
kek
this guys retarded
>>
>>108938686
>t. AI slopper

Saaaaaar
>>
>>108937901
>publication date: 2021
>AI chance: 0%
This is what is going on. It just checks WHEN it was published, not the contents.
>>
>>108937844
>contain AI prose
that's a massive difference from
>written by AI

every paper needs some bullshit sections on how it will change the world and why the funding authorities are meeting their KPIs by giving you money to work. there's also the simple fact that if you are writing a dissertation (PhD thesis) you probably have to dump and shorten a lot of material from your academic publications, which is a perfect job for AI tools. use it to get the bulk material done, then edit by hand. when I wrote my PhD thesis I had to dump like 6 papers into it.

also if you're not anglo you have to do things like translate the abstract into your native language. fuck doing that by hand.
>>
>>108937844
>contain AI prose
The AI contains human prose though
>>
>>108938946
how do you know pangram is ai retard
>>
Have they said if specific fields are more prone to this then others?
>>
DESU AI-generated texts have already caused irreversible damage. Pretty much every article, book, and thesis published after 2022 contains AI-generated garbage. If you want to find info on a subject you're interested in, you either:

1) read sources published in 2021 or earlier
2) read post-2022 sources and basically get guaranteed AI slop

People barely write texts by hand anymore. StackOverflow is dead. Quora is dead. Medium is dead (feels like it’s just pajeets churning out low-quality AI content at this point). Books are being AI-generated with "authors" pumping out dozens of them a year. There is literally nowhere left to get new, original information.
>>
>>108938998
https://www.pangram.com/research/how-it-works
>>
>>108938998
I didn't and I wasn't assuming that but apparently they tell us so lol.
Anyway, it is very hard to assign 0 or 100% confidence to something, so if it says something has 0% chance of being AI genned, it is using some kind of fixed comparator like year published.
>>
>>108937844
I have tried several tools on my thesis and it all says its AI. Thats bullshit I wrote it a decade ago I just have a no bullshit style and try to keep complex words out.
Also having AI improve the text was a massive failure it always lost the meaning and added new random bullshit. I don't get why people say its going to replace professional writers those at least maintain the central meaning of the text.
>>
>>108937844
ai detecting ai is trash
you just have to write like a intelligent person that needs bloat in their thesis and suddenly is AI flagged
reminder that every thesis is filled with bload because professors like that
>>
>>108937844
Not surprising. Academia is rotten. There is very little real research going on. Any retard can be a "researcher" nowadays and leech money off taxpayers, nonprofits, or companies paying for fake research.

It's a corrupt pyramid scheme. Surprised this 25% figure wasn't higher, it's probably a very conservative estimate. Higher ed is a scam, if you willingly choose to participate in that system you have no morals.
>>
>>108937935
If I'm smart enough to hire people to do work for me then I'm going to start a business—the opportunity cost of spending that time and money on a dissertation just isn't worth it; there's a reason people with PhDs are overwhelmingly poorer than CEOs.
>>
>>108939638
I hate writing so I've used it quite a bit to spit out a first draft based on some sort of result description document with bullet points and technical jargon. Then I can look at the first draft, get mad at all the stupid bs the AI has spat out, and end up re-writing 98% of it.
>>108937935
You're retarded for thinking any use of AI for text editing means all the underlying content and results were AI generated
>>
>>108939666
Everyone uses AI nowadays because everyone else is using AI nowadays. No human can manually pump the amount of bullshit text that you have to come up with to get any funding nowadays.
>>
>>108939580
>wasn't assuming so
that anon already proved you right, why'd you have to show how dumb you were like that?
>>
>>108937844
OPEN DATA SHOW CHART AND GRAPHINE
>>
i don't see the big deal here.
you're going to be expected to use the llms in workplaces, including research roles.
>>
>>108939690
jews?
>>
>>108938920
Good quip ranjeet! You're finally fitting in!
>>
>>108937884
that just means it is over-fitting and has lots of false negatives
>>
File: Capture.png (64 KB, 588x582)
64 KB PNG
>PhD in education
It was a joke even before AI.
OP is misleading.
>>
Maybe with the AI apocalyse coming for academia now, they should propose students make actually useful contributions to society instead of navel-gazing dissertations nobody will read.
>>
>>108939738
>You're retarded for thinking any use of AI for text editing means all the underlying content and results were AI generated
You're retarded for thinking that any use of outside labor for text editing means all the underlying content and results were written by the guy I hired
>>
>>108937844
Sounds like a surefire way to become bankrupt for life when you get sued for defamation.
You really want to end up like Alex Jones or Amber Turd?
>>
>>108941737
Yes, exactly. If you did the research and math yourself, and can DEFEND IT at your PUBLIC THESIS DEFENSE literally nobody cares whether the actual prose of your dissertation was written by you or Ernest Hemingway.
>>
>>108941717
>sample of 100
>contained SOME amount
that nigger will flag everyone who uses AI for so much as generating an abstract or chapter tie-betweens
>>
>>108937844
Now break that down by race/ethnicity. I have a feeling it's a lot of Chinese, Indian, and jewish students.
>>
>>108942692
Exactly so why do you keep saying there's anyhting wrong with hiring someone to write my dissertation so long as it's my ideas?
>>
>>108942709
not only that but Pangram, which I've tried with a throwaway email, is just like every other ai detector, BS that operates like a coin flip. I flagged as 100% human with high confidence some text I wrote then edited with AI. And I expect it to be the same the other way around. And their technical report paper (which is not peer-reviewed btw) is also a bunch of shill garbage.

tldr: the twitter tard in OP is a shill.
>>
>>108938686
IMO this kind of thing is necessary. As an American, one thing I really loved about Russian academia when I lived there was "Dissernet", which is a volunteer organization that goes through an absurd number of dissertations and publications to weed out plagiarism and ghost-written work.
Academic integrity is really important; the amount of absolute bullshit I see coming out of even supposedly reputable institutions is actually insane.
>>
>>108942778
whoah there buddy, we don't tolerate antisemitism here, haha
>>
>>108943567
it flagged my entire ai slopped 'research paper' as a human written
it's a buillshit
>>
>>108937935
Having an AI to improve organisation and help you communicate ideas is nothing like having someone write your paper. Also supervisors are lazy and retarded so with an AI you don’t have to wait for their dumbass to read and give feedback
>>
>>108943599
The problem in Western countries (mosty because of jeets, browns, chinese), is that it’s not plagiarised but it’s right on the line of being satisfactory. When everyone under a supervisor is pumping out carefully selected lazy projects, they won’t fail anyone or start expecting more. In the end AI is only helping a fast descent into even less useful outcomes from universities
>>
>>108937844
i'm pretty bad at writing prose so i wrote the raw version of my masters thesis myself and just let it run through a llm paragraph by paragraph. it mostly worked though i had to correct a bunch of stuff because the AI writing style was sometimes way too obvious

not that surprising that others also did something along the line
>>
>>108937844
>made up bullshit
>made up bullshit, written by AI
In the end, what's the difference? Cheaters gonna cheat.
>>
>>108937844
As a funny anecdote, about two months ago in my retarded uni we had to do a pre-exam activity. I didn't use AI at all, it was easy and short, no need, but the teacher's grading was the minimum passing mark plus a warning for... having used AI????
When I tried to complain she didn't believe me. The last two activities I did em entirely with ChatGPT because fuck her and for some reason she graded me excellently on both. What do you think it happened? Do I write like an NPC, a dumb retarded chatbot? Could she have used one of those crappy scam AI detectors? I live in a thirdie shithole if that matters
She never explained the reason for this aside from "it's obvious u used AI, I can tell"
>>
>>108945196
women



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.