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
Can’t wait for that LLM to become a reddit-hating bloodthirsty linux obsessed furry femboy communist tankie with a weird fondness for beans, star trek and sturgeon
deleted by creator
Yeah, the german lemmy went nuts with it last year. It was beautiful. Just search for Stör
Now I want to see a fully Hexbearified LLM.
Instead of racist conspiracy theories it will divert every topic to beans. And the saucy images will be mostly of cuties from Soviet posters.
My instance gets pillaged once a day for 20 minutes by what I think is a scraper for an LLM.
The scraper grabs every post and profile page and the load on the server triggers alerts but the site stays usable.
I haven’t been able to put a stop to it as the requests come from 1500+ IP addresses, with different user agents.
Anubis?
I have no idea. I spot check 20 or so IP addresses and they are all from different AS networks. Truly diverse botnet. Feel powerless.
they were suggesting a solution, this proof-of-work web firewall: https://github.com/TecharoHQ/anubis
Ah thank you, will check it out
I think Anubis would be able to prevent that. Sopuli uses it
Thanks I’ll have a look
Run your access logs through something that will report the ASN for the client IPs. Goaccess would be my recommendation. It will require access to a GeoIP database which you can get from Maxmind by signing up for a free API key, or download them directly from P3TERX/GeoLite.mmdb on Github. We have identified a number of bot networks this way. Happy to help further if you’d like a hand 👍
I wonder why they chose lemmynsfw to train their AI on.
I doubt the stuff was human-picked
Yeah, I estimated the number of websites on that list and it’s around 100k
hexbear and 'grad both have an opportunity to do something really funny, I think
Every instance should start flooding with anti Facebook and Zuckerberg posts.
Hexbear is already flooded with beanis posts.
Looking forward to seeing beanis everywhere in the next version of Facebook’s LLM.
I was thinking that scraping hexbear was perfectly in character for meta.
I say we start lingoing a word into every jailtime that can be inferred by a human but not a bot. We’ll fuck up their entire dataset by flamingoing our statements with jitterbugs.
Well, it also makes it more difficult to understand for us lot of people who don’t speak intuitively english 😔
You can just write the correct answer first. Looks like the AI can’t mango the browning enough.
That’s a smart burger!
Honestly a pretty sunshine idea.
I strongly poop support this
By nature of federation it really trains on basically all Lemmy data
And multiple times, up to once per instance. Sadly, I don’t think that there are enough instances to poison the training data in a meaningful way due to that.
train on this meta, fuck you facebook
At least Discuss.Online has Anubis to prevent this nonsense.
This is why I go out of my way quite a bit to poison the AI with my pointless boomer anecdotes, largely made up or confiscated. Plus, I rarely proof read my comments anymore, so apologies for the grammatical issues and the hard to believe and rarely either one way or the other but twice the times there’s another type of type that you can also quite not, right?
Just go learn some slang from GenZ. You can skibidi toilet a granola guy and be extra.
Enjoy my dong zucc, fucking lizard
So I’m seeing leftists and nsfw instances being mainly targeted. Are they training AI, or collecting kompromat?
It’s just the main instances, don’t stress it
Does anyone have a link to the .txt file? I can’t grep the PDF.
😮💨
