this post was submitted on 13 May 2026
31 points (84.4% liked)
TechTakes
2567 readers
262 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
You should follow the comments in those links showing that this is deliberate disinfo.
Here's existing discussions: https://piefed.social/c/fediverse/p/2032935/ai-assisted-moderation-in-the-fediverse-is-happening-now-what
From the horse's mouth: https://lemmy.dbzer0.com/comment/25880065
amazingly, I did actually click on those posts! and those posts did inform me that this is happening!
what part of the information do you consider wrong? where's the code? I'll go read and see where it disagrees
Oh you did read them? How did you miss the part where we pointed out that everything Rimu said was misleading and we don't actually moderate with LLMs?
did you read my post and see the bit where I don't make any moderation claims? where's the code?
The code of what? The script that uses the lemmy API to download the public post history of a user?
You linked to disinfo and them claimed we"feed posts wholesale to prompts". You know the implication you were trying to make, and as it's clear from the comments from others here that it's working.
are you being intentionally obtuse? the code to do everything involved with this issue. the code that would be a very simple demonstrative "oh, okay, so that's what's happening" read with literally no ambiguity, as is extremely usual to do in open source. that code. where is it?
This is an internal tool that a mod developed and an admin was trying out. You realize that at this point I could just post any fucking code I want to prove whatever I want, right? So you understand this doesn't prove anything? So why not just believe me when I tell you that all the script was doing was downloading a user's public post history via the API?
I anyway asked the developer to share the code with me, as again, it wasn't an official instance tooling.
The code was used to pull a users comment history into a text file, which was then used to ban the user, right? Leaving the fact that it was then fed into an LLM that may or may not have been self hosted based on "jokes" aside, how is code used in an official instance action like moderation not "official instance" tooling?
Do you post a public log of all your "official instance tooling?" How often, and in want instances, has "unofficial instance tooling" been used to do official instance actions like banning users?
then my take/read is that your team simply sucks at this
your posts across various threads about this for DAYS are all running interference, and you could simply have skipped past that by going "look, here's what ran, here's the logs"
and as I said in my other comment on the other branch: running around shouting like you are isn't a great look
why are you going to bat for someone using unofficial tooling that proved to be extremely unpopular because it was used in a manner that looked like abuse to most observers and qualifies as abuse for instances like ours? it feels kind of rich to come here angry when other people are asking you to stop. maybe you can just take our word for it that we don’t want an LLM anywhere in the moderation process, even as a post summarizer? you don’t have to understand why if you don’t want to.
and yeah you absolutely can lie about this crap. it kind of makes it worse that you and your mod felt they didn’t have to. we really don’t want any of this normalized on any level.
forgive me for I have sinned:
sins
I am not defending them. Trust me I am plenty annoyed that I have to deal with the fallout and the constant disinfo it opened us up on. But at the end of the day, we're humans and people are new at moderating and new to social media, and don't always understand how visuals come across and how easy it is for people to pick up the pitchforks, and I'd rather make this a learning moment than a purging moment.
I will absolutely accept the shit that comes our way because people didn't think how their actions would look (again, this doesn't mean that LLM for moderation is good but being seen using is bad. The former is bad too.) But I don't like being blamed for bringing the frigging LLM apocalypse to the fediverse, you know what I mean?
For the record: I'm not angry at people telling us to stop. I am angry because we never even started and I keep saying this and it feels like people are just not listening and keep repeating the same disinfo and I have to keep saying "People, this is not true, this is not at all what happened" again and again.
absolutely! please show me you’ve learned from this moment by taking a stance against the use of LLMs anywhere in the moderation process and making it part of your instance’s policies going forward.
I think that would just be performative at this point, but I'll discuss with the team.
perform away! please keep in mind that Lemmy and ActivityPub are communities, and the only proof of intent we have is through action. a change in policy might feel performative, but it gives me something I can boost as a positive example and a sign that things will be better going forward.
We're probably long overdue with an official meta post in our instance about this whole shitshow that we can link on the next time this disinfo is pushed.
you thought I’d be ok with this shit because it was done with an allegedly open source model? I’m so glad you ignored all our posts about unchecked algorithmic bias and the fucking rotten origins of every open source model (hint: you don’t have the resources to train a model because you aren’t a billion-dollar company and that’s by design) because moderating your instances appropriately was ooh just too haard
my guy what in the fuck are we doing here? more to the point what are you doing here?
You'd think you'd be able to see the posts I linked and realize we didn't actually moderate our instances with LLMs?
yeah, it’s ok because the LLM wasn’t hooked up directly to the ban API, you just used it systemically to not do the only fucking thing you’re supposed to be doing as a moderator
this human in loop shit is how corporations absolve themselves of responsibility for decisions taken purely on the word of an LLM. it lets them fire a worker instead of an executive. you’re sure this is the route you want to go?
Can you fucking read? What "systematically" are you talking about? How did we use it " only fucking thing you’re supposed to be doing as a moderator"? Is summarizing "the thing you're supposed to be doing as a moderator"? Is doing a summary once "systematically"? Why are you continuing to spread disinfo?
Once again. The admin in question DID NOT USE THE LLM TO DECIDE ON THE ADMIN ACTION. Can you understand this? Can you read this? Am I talking to a wall?! You are swallowing disinfo and then spitting your outrage mindlessly on people.
Literally no one believes this corpo speak bullshit. That they just coincidentally ran this unpublished python tool, did their own work, then just happened to use an LLM to do the exact same work right after, totally innocently? That reads as absolute ass covering and nothing more. This is the "I smelled weed" of cop stops, just filtered through nerdy fediverse bullshit.
Then, because the above totally happened like you said it did, as a one off joke that no one would ever notice, the same admin opted to put a current OpenAI model name in the LLM field in an absolutely not tongue in cheek way for other admins to totally catch and joke about? Which of course happened, haha, y'all had a big laugh about it before this blew up, yeah?
Oh, and of course this only happened the one time, and never again. Of course no one on your team used this unpublished time saving and thought terminating tool again, of course not.
Come the fuck on.
lol, lmao even
you do realize all this comes across as running interference while feeling real guilty? you could just, like, drop a link to a repo with the bits. you know, with commit history. like i've been asking you for. just simply linking to that would be a real quick way to settle the issue! it's even some free code review for you!
You realize there might not be a repo, right?
you realize that mod probably shouldn't be one, right?
seriously this just looks more and more embarassing the more you post about it
As I said elsewhere, I'd rather this be a learning moment, than a purging moment.
the two things can go together, and not removing said mod means that if they do not learn they can still do plenty enough damage elsewhere later on
and I say this out of the position of having fucked up in a mod role before
no but you’re about to be talking to a defed unless you stop this all caps fake outrage shit. should I do you the favor of deleting the garbage post you just made so you can give it another go?
think very carefully on this because I’m not sure pretending to not understand what I’m asking is working out for you
assume I know and understand that the LLM did not literally do the banning
maybe there’s a large body of existing research on how even human in the loop systems confirm and worsen biases? maybe it’s a bit obvious when you go through the process the moderator took to get to their decisions in your head? slowly now, maybe you’ll get there
I am telling, again, that the human did not use the LLM to think for them either. The admin took the decision to ban the user irrespective of the LLM, and the rest of our admin team and me specifically, would never let an admin become a "human in the loop". The LLM was used just to summarize, as part of the test, with a misguided inside joke on using OpenAI tech.
I will readily admit that there was mistakes made by the admin. Not on their actions, but on their visuals. Because those visuals were spun to keep feeding this made-up controversy. We didn't use the LLM to decide or even guide our decision, but it appeared like we did, and we already owned that.
you don’t appear to have much understanding of how a human in the loop system works in practice. LLM summaries are used to confirm biases, especially when the prompt is something along the lines of “do these posts contain ?” though these systems are stochastic so you’re going to get unpredictable biases regardless of the prompt.
I don’t accept that the LLM summary didn’t influence the decision because the mod in question confirmed that he knew the LLM agreed with him (that’s bias, and also not something LLMs are capable of actually doing) and because if it didn’t, then the summary is worthless
which is why maybe you should just not have them in the future? just don’t touch LLMs when you’re doing mod work. either there’s no reason for it or you’re doing something monstrously wrong.
In this case, according to the admin in question, the LLM summary came after the decision, as a sort of a test. I.e. the admin made a decision, and wanted to see if an LLM would subsequently agree with that decision. In this specific case, it did, which is why they misguidedly decided to keep its summary in the modlog (opening us up to this whole shitstorm), but ultimately, that admin anyway decided LLMs in the mix is not good at all, which is why you never again saw an LLM summary in the modlog.
I can only put so much fault for a person for just testing shit out, yanno? I am not happy that they decided to use the output of the test because they are not familiar with how quickly disinfo breeds, but ultimately they came to the right decision anyway. If they had not and they had raised the issue on using LLMs officially, they would have been shut down.
Having a LLM confirm a decision is the same thing as having the LLM make a decision and then figure out if the mod agrees with it. If they would have chosen not to rule based on the LLM output, then the LLM was part of the decision making process. The order does not matter.
Including the LLM outputting something that implies a determination at any step automatically makes it part of the process.
hey fucko, you know we don’t have to take their word for it right? we can read all the relevant posts and come to the conclusion that actually the use of LLMs as stated fucking sucks, and that we don’t fucking want it. we can read something and come to a different conclusion than you, believe it or not.
I completely agree with you.
Here's the answer to the only question you posted which should be obvious from everything else I've said and done.
cool! please make it clear to the mod in question that they shouldn’t be using an LLM for anything in the future, even summarizing posts. make it part of your instance’s policies.
How could an admin possibly enforce that? What if a mod created a summary locally and never shared it with anyone? The ai summary wasn't used as evidence, that is already policy and has been explained to you multiple times. You are shifting the goal posts to the moon, and no policy change will ever satisfy you.
oh no the goalposts! think of the theoretical shitheads who might do this in secret!
please see the pinned post and fuck off