Dropsitenews published a list of websites Facebook uses to train its AI on. Multiple Lemmy instances are on the list as noticed by user BlueAEther
Hexbear is on there too. Also Facebook is very interested in people uploading their massive dongs to lemmynsfw.
Full article here.
Link to the full leaked list download: Meta leaked list pdf
Meta AI’s gonna go dong out tankie
Well, obviously. It’s an open protocol. I assumed everyone on the Fediverse must be pro-AI training, otherwise why post here? That would be dumb.
Uh… Are you saying simply using social media is endorsing stealing personal information to train LLMs? Because that’s a wild take, if so. Personally, I feel like there’s no stopping them, so what am I supposed to do, stay silent? Not engage with anything, even anonymously?
Uh… Are you saying simply using social media is endorsing stealing personal information to train LLMs?
Of course not. No “stealing” is happening. People are posting content on an open protocol that permits anyone to read it. Exactly as intended.
Personally, I feel like there’s no stopping them, so what am I supposed to do, stay silent?
If you do not want to be heard then yes, I suppose you could stay silent. That would indeed accomplish that.
You could also find a social media platform whose content is locked behind a walled garden of some sort that makes it more difficult for your posts to be seen by the public. But that’s antithetical to how the Fediverse works, you want someplace very different from here if that’s how you want to approach this.
Basically, you are on a platform that’s specifically designed to broadcast your comments far and wide without restriction, and then you’re getting upset that someone you didn’t want to hear your comments is hearing your comments. I’m not sure what you expected.
Participating in a public forum that has no technical way of preventing data from being used by a particular class of actor does not preclude having an opinion that a particular class of actor should have rules about what data they are allowed to use.
People can have whatever opinions they want to have. In this case that opinion flies in the face of obvious reality and I’m pointing that out.
It’s like trying to drive your car across the Atlantic ocean and then griping about how the car failed to stay above the water because you really thought it should be able to handle that.
It doesn’t matter how many pithy analogies you make. You need to recognize the difference between “I know they’re scraping this website because they can” and “I don’t think they should be allowed to scrape this website”. You’re arguing that they’re incompatible when they’re not.
As I said, people can have whatever opinion they want. Reality is under no obligation to respect those opinions.
Analogies are merely explanatory.
If you understand, then you should be able to understand that your “they were dressed like they wanted it” level argument bullshit is completely unnecessary.
It’s no different from posting on usenet back in the day. You have to assume it’s all completely public. People are gonna train their AIs on large public text repositories whether they’re given permission to or not.
Fuck a Zuck.
Can’t wait for meta chatbot to tell zuckerberg to kill himself
I mean, think about it.
The majority of written communication are core focused on Drama.
People communicate online about subjects --> Drama News - Drama Books/Stories - Drama
So is anyone really surprised that you train an entity on only written text and it ends up being dramatic?
Sure, this is open data viewable by everyone.
Stands to reason that AI is being trained on it.
So, duplicating their data? That seems counter-productive.
It seems counter productive for them to scrape it when the API is right there
I remember back then, some people defended not blocking Threads instances.
And one of those defenses was “it doesn’t matter if you block Threads, the underlying ActivityPub protocol is open and anyone who wants the data can still receive it.”
Turns out to be the case. It didn’t matter if you blocked Threads.
I think a better reason is the federation only worked one way. Why should we share our content if they’re not sharing theirs? Not that we’d want it.
Seriously? Meta uses many methods to snoop on the cell phone and with its functions it also looks for devices in the network in which you are logged in and also devices simply in the vicinity. It goes without saying that Meta makes use of open data… I would even go so far as to say that other AI models are not trained any differently. Well, they may be trained using an AI that has been trained on them so that they don’t have to access the data from the actual sources themselves.
Oh ya? Suck the cuck should get a dildo up his arse.
Is there an easy way to poison the input? Is there something we can slip in our comments that could make the data useless?
lmao wtf is that list. Literally training their AI on beastiality.
Edit in case it’s not obvious: That domain is very much NSFW and it’s exactly what you’d expect (I checked and wish I hadn’t).
hmmm…
