We are defederating awful.systems from:
because it’s extremely anarchist to use the chatbot as your moderator (no no it’s totally human in the loop) and post corporate DARVO-filled excuses when caught.
This is a last straw after a string of dubious and exhausting behaviour.
If you’re an awful.systems user and you post to communities on those instances, you will need to create an account on one of them.
Man they wrote a lot of words to say “We didn’t use an LLM to feed it user data, except for the fact we did.” Cool distinction that it wasn’t used to decide on the ban, just to communicate it 🙄
Happy for you/sorry that happened
I don’t have anything especially useful or insightful to add, but I just want to thank our mods for acting decisively to defederate us.
Edit: seeing the discussion over on https://awful.systems/post/8293548 I think our mods were actually extraordinarily patient in giving db0 a pretty solid chance to own up to their mistakes and commit to no LLM usage in the future and they instead doubled down hard.
db0’s roundup doesn’t pass the sniff test either. i’m supposed to take on your word that it was an “experiment” (testing what hypothesis?) and an “inside joke” (where is the explanation?) despite the paper trail showing approximately 0% of this being true, but you’re totally comfortable speculating on the actions and motivations of Rimu – nay, a mysterious unnameable “piefed developer”? you have the gall to be salty they didn’t edit your narrative into their post about your instance admin’s behavior when you refuse to even say their fucking name? it’s gross! it’s gross.
lol at “corporate-LLMs”
after scrolling through a couple days and accounts on the extended db0verse it’s extremely hard not to come to the conclusion that it’s a haven for people who think Rimu specifically is some kind of combination antichrist/lolcow who is personally here to destroy all of mastodon, lemmy, and the concept of federation itself. some choice extracts from the very fine people over there:
[Rimu] is an active danger to the entire fediverse as a whole. […] you better expect questions on why you’re helping to further this narcissistic psychopath['s] control and influence. […from that user’s bio:] Rimu is weaponizing the fediverse against you.
very reasonable and normal account to have on your instance
Rimu is breaking the Fediverse. We had a good thing going for a few years.
that’s right, he’s personally deleting every post on the internet forever
Hi. Admin of Lemy.lol here. […] When I joined the Lemy.lol team, I was informed that Rimu had pre-emptively defederated from us sometime ago due to our logo. Pepe the frog. [… He’s] not merely a [power-tripping bastard], but a dangerous one.
fucking lol it’s these chuds again
Did Rimu suffer a TBI or get way into something mind altering?.
that last one is on a thread completely detached from Rimu, by the way. Rimu possesses a kind of hyperagency, capable of superconvincing everyone of everything which is why any and all defederation due to AI slop is actually secretly his fault
so yeah, good riddance
also, poor wittle intern mod unruffled who was there as an admin barely for three years since day one, that’s no time to learn how to moderate at all
then after defed immediately came complaining on dedicated community
I could be charitable enough to imagine a conversation on Discord where one mod posts “heres the log, ngl he looks pretty bigoted, i already pre emptively banned them”, next mod posts “lol ran it through deepseek for funsies and got <clipboard.png>”, and finally somebody says “lmao gettin smarter erry day, verra nice, big thanks to big brother sponsor OpenAI™”, to which there are many Borat stickers and thumbs-up. I can’t be charitable enough to ignore their own wiki:
Divisions by zero is a lemmy instance hosted under the dbzer0 domain and is a founding member of the FAF. It is run by Anarchists but is not exclusionary to the ideology. It promotes libertarian socialist ideals, is against copyrights and pro-non-corporate GenAI technology. It is neurodivergence-friendly.
Wrong fucking dog-whistle, dudes. If you want to communicate that you’re open to responsible self-hosted machine learning which doesn’t simulate humans, then say that. As read, this sounds like they enjoy using
llama.cppat home, and the median friend in my orbit who is doingllama.cppat home is suffering light chatbot psychoses already. Honestly, it’s a little cringe that they haven’t declared aNeo-LeninistAnarchist Butlerian Jihad By Zero instead.libertarian socialist ideals
wtf are these ideals supposed to be?
If its “libertarian, but doesn’t want people dying in the streets or dying of preventable diseases”, I will give them the tiniest modicum of credit for being better than standard libertarians, but being libertarian at all still leaves them deep in the negative on credibility in my books.
pro-non-corporate GenAI technology
By which I assume they mean corporate models that got released as open weight models (and at-best/at-most got a touch of fine-tuning from a community effort), but still ultimately originated from mass plagiarism and are still not useful beyond generating slop…
“libertarian” outside of a US context is a perfectly valid word to use to describe “non-authoritarian” or anarchist left ideologies, e.g. libertarian communism or “libcom”
i suppose you’d say “left-libertarian” if the meaning is ambiguous, but here it should be implied by “socialism”
Very little relation to capital-L Libertarians. Libertarian socialism is broadly socialism that seeks to replace the state with smaller-scale social structures and direct democracy that can satisfy everyone’s needs equitably without the threat of violence or enforcement of property rights. Libertarianism is meant to protect capitalism from the state; libertarian socialism is meant to protect people from both the state and capitalism. I don’t want to get into the details of leftist ideological micro-labels, but generally confusing anarchists with anarcho-capitalists is the quickest way to piss off both.
lol if i’d seen that i’d have moved to part ways long before this, goodness me
Ray Kurzweil Thought gonna be stimulating ideological incoherence for years to come
your read of this situation is entirely correct, but things aren’t as bad as you think. things are a bit worse
they’re running a distributed computing thing that hosts some llms and diffusion models, for their own and their users use
and, frankly, given the “what makes you think there even is a repo?” type answers, I’d have some Ops Maturity questions there too
(nvm even the code itself (which apparently is something in the ballpark of heisenberg dynamics))
update 2: they found the repo, week+ later, but won’t show you and it was made by yet another admin (tbf it was said straight previously deep in some thread). just a rogue mod, nothing to see here
i’m not a professional computer toucher, but i bet you’ll find something to sneer at here https://haidra.net/deleting-users/
I kept asking for the code, and their response was all manner of telling
Good on you for pushing that point. Trying to pass off the logging details identifying their llm as ChatGPT as a “joke” was incredibly stupid, but more people might have bought that lie if you didn’t press that followup question.
Yeah I’d like to see the python “query” script and its git history to ensure it just pulls user comment into a txt file and doesnt then autoprompt it into an LLM api/mcp. That’s the logical end point of querying user comments if you’re going to use an LLM to “analyze” them anyway.
Maybe some proof of this qwen instance with setup logs or something that shows it existed at the time of the supposed local query.
This right here.
LLM usage is bad, but could be forgiven if the instances shut it down and put a plan in place to prevent a recurrence. A little “sorry” goes a long way.
The lying means that they aren’t sorry and can’t be trusted.
we’re talking about openly pro-genai instance where good majority of community banners is slop and that runs distributed chatbot cluster. apparently when you don’t just enourage brainrot as a service but also take it into your own hands and cook your own users autonomously, that must be some kind of revolutionary praxis
given opportunity they’ll make anarchist palantir then will insist it’s all fine because it’s open source, sorry “open weight”
Hmm, a good move, as the implications using LLMs in moderation work is a big no-no. As moderation requires both human interpretation and care in order to be successful. Feeding the latest torment nexus that the Silicon techno-fascists have concocted with user comments is only doing long term harm. While there is no money in generating slop for the masses, there is in surveillance, and feeding user comments to the beast is only going to help those twats. I’ll have to figure out my instance’s position on LLMs and see if I can get them to defed from these three offenders. If not, I’ll happily spin up an account on an instance that defederates from the growing list of ick: lemmy.ml, hexbear, and the db0 triplets…
If you’re an awful.systems user and you post to communities on those instances, you will need to create an account on one of them.
Good thing I haven’t heard of either of them until now.
the only worthwhile thing they do is piracy and they’re very proud to be the only worthwhile place for it
the awful dot systems open source moderation model codenamed anarchy cop spit out this image:
image description
kids on a police tricycle
what does it mean?
sorry sorry the chatbot posted a fisher price police tricycle im trying to delete it
(image deletion is a lemmy joke for the advanced sufferer)
Removed by mod
We are aware that our instance is seen as “LLM-friendly” due to our nuanced take on LLMs
AI-haters are making themselves irrelevant so why care. Using local LLMs for various tasks they’re well suited for is just being smart.
wow I wonder why having you around is fucking exhausting
cultivate this horseshit elsewhere thanks
don’t you know, nuance is when you leave it to chinese confabulation engine to determine political leanings
chinese confabulation engine
The best case scenario (for them) is that is a Chinese confabulation engine. But that requires that the logs mention ChatGPT as an in-joke. And code or logs of interfacing with a local instance of the Chinese confabulation are for some reason not available for totally innocent reasons. The more likely scenario is that it was ChatGPT.
I was trying to figure out how to make that post earlier
(I know that range has been studied a bit but I’m not up to the latest on how much, and I ain’t gonna go post out my ass)
