- cross-posted to:
- pulse_of_truth@infosec.pub
- cross-posted to:
- pulse_of_truth@infosec.pub
Wikipedia has a new initiative called WikiProject AI Cleanup. It is a task force of volunteers currently combing through Wikipedia articles, editing or removing false information that appears to have been posted by people using generative AI.
Ilyas Lebleu, a founding member of the cleanup crew, told 404 Media that the crisis began when Wikipedia editors and users began seeing passages that were unmistakably written by a chatbot of some kind.
You must log in or register to comment.
Further proof that humanity neither deserves nor is capable of having nice things.
Who would set up an AI bot to shit all over the one remaining useful thing on the Internet, and why?
I’m sure the answer is either ‘for the lulz’ or ‘late-stage capitalism’, but still: historically humans aren’t usually burning down libraries on purpose.
State actors could be interested in doing that. Same with the internet archive attacks.
historically humans aren’t usually burning down libraries on purpose.
How on earth have you come to this conclusion.
To be fair, it’s usually to effect cultural genocide. It’s not average people burning libraries, it’s usually some kind of authoritarian regime.
* looks around and gestures broadly in agreement*
Florida says hello. A bunch of other places too, sadly:-(.
historically humans aren’t usually burning down libraries on purpose.
Sometimes they are, Baghdad springs to mind, I’m sure there are other examples. And this library is online so there’s less chance of getting caught with a can of petrol and a box of matches.
Then there’s every authoritarian regime that tries to ban or burn specific types of books. What we’re seeing here could be more like that - an attempt to muddy the waters or introduce misinformation on certain topics.
It’s not about on purpose but usually most people don’t care about what’s not in their interest. Today interests are usually quite shallow what tiktok shows quite well. Libraries do require money for operating. Even internet archive and wikipedia
People like this
Yeah but the other thing about humanity is it’s mostly harmless. Edits can be reverted, articles can be locked. Wikipedia will be fine.
Edits can be reverted, articles can be locked.
Sure, but the vandalism has to be identified first. And that takes time and effort.
Wikipedia relies on sources, and humans choosing the sources like newspapers. And those newspapers are more and more inside a “bubble” that rejects any evidence or reporting presented by a competing bubble.
Right now wikipedia is covering up one of the greatest acts of mass murder of our times, because the newspapers are covering it up, or rejecting evidence because it’s by the “enemy”. Part of this is a defensive posture against AI bots and enemy disinformation.
Maybe a strange way of activism that is trying to poison new AI models 🤔
Which would not work, since all tech giants have already archived preAI internet
Ah, so the AI version of the chewbacca defense.
I have to wonder if intentionally shitting on LLMs with plausible nonsense is effective.
Like, you watch for certain user agents and change what data you actually send the bot vs what a real human might see.
I suspect it would be difficult to generate enough data to intentionally change a dataset. There are certainly little holes, like the glue pizza thing, but finding and exploiting them would be difficult and noticing you and blocking you as a data source would be easy.
I have to wonder if intentionally shitting on LLMs with plausible nonsense is effective.
I don’t think so. The volume of data is too large for it to make much of a difference, and a scraper can just mimic a human user agent and work that way.
You’d have to change so much data consistently across so many different places that it would be near-impossible for a single human effort.
I never told that I think it is smart…
Its because there’s no accountability for cybercrimes. If humans always had a button to burn down libraries, I’m sure they would have. Instead they had to put themselves in harms way to do such things.
People do things cause they can, and fucking with Wikipedia is apparently simple.
As for why this is happening, the cleanup crew thinks there are three primary reasons.
“[The] main reasons that motivate editors to add AI-generated content: self-promotion, deliberate hoaxing, and being misinformed into thinking that the generated content is accurate and constructive,”
That last one. Ouch.
The vast majority of people think they’re the good guys…
Are we the baddies?
Without knowing you: probably.
Russian agriculture is in dire need of mechanization.
“[The] main reasons that motivate editors to add AI-generated content: self-promotion, deliberate hoaxing, and being misinformed into thinking that the generated content is accurate and constructive,”
I think the main driver behind people misinformed about AI content comes from the fact that outside of tech people, most have no idea that AI will:
-
100% make up answers to things it doesn’t know because either the sample size of data they have ingested was to small or was bad. And it will do this with the same robot confidence you get for any other answer.
-
AI that has been fed to much other AI generated content will begin to “hallucinate” and give some wild outputs, very similar to humans suffering from schizophrenia. And again these answers will be given as “fact” with the same robotic confidence.
And then #2 will be copied by other people and AIs, becoming seen as fact.
And then #2 will be copied by other people and AIs, becoming seen as fact.
-
Well, I was in doubt, so I asked the AI whether I could trust the answers and it told me not to worry about it. That must mean that I only get accurate answers, right? /s
Unleashing generative AI on the world was basically the information equivalent of jumping headfirst into Kessler Syndrome.
For the uninitiated like me:
The Kessler syndrome (also called the Kessler effect,[1][2] collisional cascading, or ablation cascade), proposed by NASA scientists Donald J. Kessler and Burton G. Cour-Palais in 1978, is a scenario in which the density of objects in low Earth orbit (LEO) due to space pollution is numerous enough that collisions between objects could cause a cascade in which each collision generates space debris that increases the likelihood of further collisions.
Good call, thank you.
Also: Referencing Wikipedia in this context is kinda funny.
I did think that. :) It’s just… So good. I hope it never enshitifies. God help us.
Best case is that the model used to generate this content was originally trained by data from Wikipedia so it “just” generates a worse, hallucinated “variant” of the original information. Goes to show how stupid this idea is.
Imagine this in a loop: AI trained by Wikipedia that then alters content on Wikipedia, which in turn gets picked up by the next model trained. It would just get worse and worse, similar to how converting the same video over and over again yields continuously worse results.
See also: model collapse
(Which is more or less just regression towards the mean with more steps)
Eventually every article just reads “Delve delve delve delve delve delve delve.”
A very similar situation to that analysed in this paper that was recently published. The quality of what is generated degrades significantly.
Although they mostly investigate replacing the data with ai generated data in each step, so I doubt the effect will be as pronounced in practice. Human writing will still be included and even curation of ai generated text by people can skew the distribution of the training data (as the process by these editors would inevitably do, as reasonable text could get through the cracks.)
AI model makers are very well aware of this and there is a move from ingesting everything to curating datasets more aggressively. Data prep is something many upstarts have no idea is critical, but everyone is learning about, sometimes the hard way.
Every article would end up being the philosophy page.
Sabotage Wikipedia, Ddos the Internet Archive. Makes you wonder if in the future we’re going to forget our past. Will actual history be obscured in a sea of alternative histories unrecognizably presented as the same thing. Maybe we need to keep some books laying around in archives just to be sure.
The digital dark age will be a real thing, absolutely.
Interesting idea on a sea of alternative histories. That might be a possible threat.
Someone else here called it “AI text apocalypse”. I like that term.We have still Anna’s archive, scihub, libgen and old fashion traditional libraries ( including the national ). National libraries won’t disappear in the nearest years, maybe will rotten due to defunding but still they will exist
If anyone can survive the AI text apocalypse, it is wikipedia. They have been fending off and regulating article writing bots since someone coded up a US town article writer from the 2000 census (not the 2010 or 2020 census, the 2000 census. This bot was writing wikipedia articles in 2003)
Hopefully they tightened things up after the Scots incident.
Well, for everything except fictional articles. Thats the hardest for them, historically
I hate to post because I have loved and trusted Wikipedia for years, but the fact that there are folks out there who equally trust what AI tools generate just baffles me.
The signal to noise ratio is so low these days. There’s so much information out there but everyone wants to profit from you before you can get it. Even worse, the people with good information generally can’t buy as big a megaphone as the people who profit from lying to you.
Honestly, I think humans have been more likely to believe an easy lie than a hard truth all along, but it’s easier than ever these days.
Download the torrent for the local copy of wikipedia from 2024 now
Or download it in a bunch of other ways directly from Wikipedia.
why the fuck would anyone stick ai shit on wikipedia that doesn’t make any sense
“[The] main reasons that motivate editors to add AI-generated content: self-promotion, deliberate hoaxing, and being misinformed into thinking that the generated content is accurate and constructive,” Lebleu said.
so, stupidity basically. they’re just stupid.
People suck
The irony being a huge amount of the llm knowledge was based on WP in the first place, that and scientific papers.
Don’t worry, it’s not as bad as the title suggests. The attack on Internet Archive is far, far worse. It’s obviously a bit of a problem, though.
We what? WE WHAT?!
oops
(I accidentally added an extra “we” to the end at first)
fights back by posting human-generated nonsense
StarshipTroopers-ImDoingMyPart.tar.gz
This is why we can’t have nice things
We still have it, and it’s quite nice.
Damn Putin, you’re retarded.
the name is Altman.
Dead Internet Theory all over again.
Oh God. This is horrible news. So incredibly frustrating to hear.
