Post #2595755

2026-05-14 03:40 UTC

Many past fedi firestorms have been grounded in spin and half-truths! I'd love to think that folks have become more discerning, but as you point out, no, at least on the microblogging side (which is where most of the flareups have historically happened) there isn't actually much awareness of the context. My guess is that most Mastodonians didn't know about the situation and the ones who did saw this as something that doesn't really affect them. But who knows ... In any case there certainly is a lot of context here. Focusing only on the AI in moderation, reading the followup threads it seems to me that most people believe the claims that it was local models rather than ChatGPT, and that AI was only used to confirm the mod's judgment and produce the modlog entry. In that case it comes down to how people view "AI". If "AI" a "normal technology", well, then this is just like any other potentially-biased technique that a moderator uses to get confirmation of their judgment. After all, googling for "is X a zionist talking point" is also potentially biased; for that matter so is asking some other moderator (who for all you know might have bad judgement or even be a closet zionist). if "AI" is inherentily fascistic in ways that aren't true of tech in general, then how is its use compatible with anarchist philosophy? Obviously, opinions differ! @db0@hachyderm.io @laurenshof@indieweb.social @ophiocephalic@kolektiva.social @smallpatatas@gotosocial.patatas.ca @onepict@chaos.social

Replies (2)

@jdp23@neuromatch.social 2026-05-14 06:06

Also (while @db0@hachyderm.io knows this), for those who haven't been tracking closely, here's some of the context. Some of these were also cross-posted elsewhere, and there's no doubt plenty of others. Feddit.org pushes FALSIFIED screenshot portraying dbzer0 as Nazis, bans users for questioning it. The truth is revealed.. This was from a couple of weeks ago, and highlights some of the tensions in the threadiverse involving feddit.org, Zionism, and attacks on dbzer0. I haven't dug into this myself but everybody in the thread sees pretty convinced the image was faked. The since-deleted Which instances have the most ban-happy moderators? Analysis inside, posted by Rimu of Piefed. This was posted just a couple days before Rimu posted the ":AI moderation" post. IMHO the analysis here was quite very weak, and the framing of painting dbzer0 as "ban-happy" is problematic (as well as not particularly borne out by the evidence presented, but even if it were, it's not like banning a lot of accounts is necessarily a bad thing). PieFed Flagship Instance Shadowbanning Instances from Discoverability & Other Questionable Updates is just in the last couple of days, and highlights some of the overall tensions in philosophy. what the everLLM-ing fuck is going on and Defederation of db0 instances for LLM moderation from awful.systems have a bunch of additional discussion On the recent LLM brouhaha and the FAF stance on LLMs for fediverse moderation Rimu Mass Bans Users EDIT: updated 5/15, with additional links @laurenshof@indieweb.social @ophiocephalic@kolektiva.social @smallpatatas@gotosocial.patatas.ca @onepict@chaos.social

Open ##2595756
@db0@hachyderm.io 2026-05-14 09:45

@jdp23@neuromatch.social Small correction, the LLM was not even to "get confirmation on their judgement". The decision had already been locked-in. The admin was experimenting with using an LLM to summarize the user's position, so as to improve modlog readability. But even that use case was rejected.

Open ##2595759