The 500 Hats of LambdaCalculusDear @wikipedia...
DO NOT... I repeat!... DO!! FUCKING!! NOT!! Put AI into Wikipedia!!
Don't make me get the hose!!
Claude@lambdacalculus @wikipedia what's happening?
0x150@lambdacalculus @switch @wikipedia "Helping editors share local perspectives or context by automating the translation and adaptation of common topics" oh my fucking god
Matilda Love@lambdacalculus it's vague, which means it might not be as bad as my kneejerk response thinks. but also, it's vague, which means they're probably trying to hide something.
HyperSoop 
@lambdacalculus @wikipedia yeah hey wikipedia can you please not compromise on your values. you must not allow slop generators to influence the site in any capacity
Leeloo@soop @lambdacalculus @wikipedia
I guess they saw the writing on the (Whitehouse) wall.
Let corporate fascists in, or else.
Wikipedia@soop @lambdacalculus the use of LLMs in generating article content has been opposed on many projects (notably the English Wikipedia) for quite some time, and I* don't foresee that changing any time soon.
This blog post is pretty vague, and I* wish there was more clarity on the plans. People are already pushing back for clarity within the community.
[*Sammy, one of the volunteers who runs this account]
Mark Eichin@wikipedia
Much of that policy page would be less contentious (and more precise) if it said NLP instead of AI (it's another "lost term" like "hacker", after all.) Whether the focus that brings to the rest of it is what you're going for is another question...
@soop @lambdacalculus
Much of that policy page would be less contentious (and more precise) if it said NLP instead of AI (it's another "lost term" like "hacker", after all.) Whether the focus that brings to the rest of it is what you're going for is another question...
@soop @lambdacalculus
Lauralee Dukeshire
Cassandrich@eichin @wikipedia @soop @lambdacalculus "AI" is not a "lost term like hacker" because (1) hacker isn't lost, and (2) "AI" never had a positive meaning. It was always scams. Going back to the 70s or earlier. This is like the 4th or 5th wave.
@dalias @eichin @wikipedia @lambdacalculus in the case of hacker it really just depends on who you ask
Sami Koskinen@dalias @eichin @wikipedia @soop @lambdacalculus it’s not just scams. It’s just that established AI applications are now taken for granted. Shortest route for driving? Used to be an AI problem around 2000. Machine vision for robot vacuums etc? Definitely AI as of 2010.
AI winters are real and scam artists gather around LLMs right now, but it’s not all crap. And there are other machine learning applications.
@sampe @eichin @wikipedia @soop @lambdacalculus None of those were marketed as "AI" to begin with. They were real concrete problems with their own names. "AI" is always the name for when you want folks to believe a computer can do things it can't actually do.
sodi, mallory, mira, olivia, alex, em, luna, etc
Chasalin@lambdacalculus @wikipedia luckily they don't want to (yet), as they boosted this post:
Raven Luni@lambdacalculus @wikipedia Wikipedia is pretentious AF an yway - more about looking pretty with fancy formulae instead of relatable information
smeg@lambdacalculus @wikipedia My edits citing sources get reverted for "original research" but we'll allow an LLM to generate garbage??
lianna@lambdacalculus @wikipedia For anyone who actually wants to read what this is about; it's about this new strategy by the Wikimedia Foundation:
https://meta.wikimedia.org/wiki/Strategy/Multigenerational/Artificial_intelligence_for_editors
TL;DR They're not putting LLM generated text into Wikipedia anywhere. They're thinking about maybe in the future providing optional tools to help relieve editors and moderators of repetitive bureaucratic management tasks such as automatic vandalism detection or putting a readability/reading level score.
It's a big nothingburger, TBH.
Rexo the Trans Bus 🏳️🌈🏳️⚧️🚌@lambdacalculus @wikipedia @lianna either they're just copying existing tooling or there's more to this than that because uh. we already have bots that detect/revert vandalism and a general article quality ratings system (though not necessarily a readability score)
@lambdacalculus @wikipedia @lianna I will say the "automatic translation" bit does leave me a bit concerned, but otoh afaik it already kinda exists? and like. I feel like there would be significant pushback by the community to incorporating AI-generated content in ways that would be generally considered problematic
:ulfr:
Old Stone Man@wikipedia @lambdacalculus I second that motion!
Nazo@lambdacalculus @wikipedia Wait, do they not actually get that LLMs would only subtract from Wikipedia, not add to it?
LLMs have their uses. This is not one of them.
PS. LLMs are not AI and never ever will be.
Bill, organizer of stuff
broodlyparaphrasing "we're going to use AI to automate some tasks"
but AI isn't like a bunch of scripts, it's a fucking slippery slope. don't do it. make your organization and scripting better, optimize your workflows, nobody here _needs_ AI to solve problems because most of those problems aren't actually technical, they're communication problems. __and AI fucking sucks at communicating.__
Victor Oxyrhynchus@lambdacalculus @wikipedia It axes the AI or it gets the hose.
Michael Graaf@lambdacalculus @wikipedia We Wikipedia editors have been using machine-assisted translation for about a decade, but as the article says, only to make us more productive. It varies from language to language, but generally the first draft of a machine translation is refused until a human has sub-edited it. The scare heading of "A.I." is just that.