I think the how is the most interesting part here.
The solution Wikipedians came up with is to allow the speedy deletion of clearly AI-generated articles that broadly meet two conditions. The first is if the article includes “communication intended for the user.” This refers to language in the article that is clearly an LLM responding to a user prompt, like "Here is your Wikipedia article on…,” “Up to my last training update …,” and "as a large language model.” This is a clear tell that the article was generated by an LLM, and a method we’ve previously used to identify AI-generated social media posts and scientific papers.
The other condition that would make an AI-generated article eligible for speedy deletion is if its citations are clearly wrong, another type of error LLMs are prone to. This can include both the inclusion of external links for books, articles, or scientific papers that don’t exist and don’t resolve, or links that lead to completely unrelated content. Wikipedia’s new policy gives the example of “a paper on a beetle species being cited for a computer science article.”
Speedy deletion is for deletions that require zero discussion, so it needs to be very simple and clear. For less sloppy genai there may need to be a discussion (unless it falls under different speedy deletion criteria.
Sometimes those discussions are very straightforward, but they allow for dissenting voices. But for “almost obvious” cases not a lot of effort is spent on them.
Of course. I also hope this will stop like 99% of the skiddie spam. I’m just afraid that, like it has happened with hacking in general, a noob installing Kali will get a ton of one-click ways to bypass these measures… and then, what’s next?
Genai inserting watermarking would be great, but that’s hard to do with text, in any way that isn’t easily removed.
I think the how is the most interesting part here.
Sounds fair. Only issue might be… that creating an automated cleanup tool to remove those triggers, wouldn’t be all that difficult.
Speedy deletion is for deletions that require zero discussion, so it needs to be very simple and clear. For less sloppy genai there may need to be a discussion (unless it falls under different speedy deletion criteria.
Sometimes those discussions are very straightforward, but they allow for dissenting voices. But for “almost obvious” cases not a lot of effort is spent on them.
Of course. I also hope this will stop like 99% of the skiddie spam. I’m just afraid that, like it has happened with hacking in general, a noob installing Kali will get a ton of one-click ways to bypass these measures… and then, what’s next?
Genai inserting watermarking would be great, but that’s hard to do with text, in any way that isn’t easily removed.
JHFC