Out of the loop

15175 readers

91 users here now

A community that helps people stay up to date with things going on.

founded 3 years ago

MODERATORS

zachimusprime44@lemmy.world

Patnou@lemmy.world

190

What is it with the flood of brand new accounts coming here only to self-promote their vibe-coded slop on c/SelfHosted? (dubvee.org)

submitted 2 days ago by ptz@dubvee.org to c/outoftheloop@lemmy.world

52 comments fedilink hide all child comments

Title. If this were Reddit, I could at least see it from the angle of a large audience. But the Fediverse is far too small for that. It's like every other day there's some 1 hour old account posting their slop-coded crap to c/SelfHosted.

Like, yes, brand new internet rando, I'll totally install your vibe-coded slop that makes wild claims about being a super secure messenger or whatever grand claim you're making. I actually might it it was posted by someone with a positive history here, but these are brand new accounts seemingly unconnected to anyone otherwise active on this platform.

Is it just a "flood the zone" strategy on their part? I'm not active anywhere else anymore, so maybe they're flinging their slop all over and I only notice it here.

you are viewing a single comment's thread
view the rest of the comments

[–] Xylight@lemdro.id 4 points 1 day ago* (last edited 1 day ago) (1 children)

i doubt that Lemmy is being intentionally scraped by AI companies, otherwise it'd give their LLMs even more severe brain damage.

[–] CombatWombat@feddit.online 2 points 1 day ago (1 children)

It's hard to find datasets on the internet that are exclusively human. You can fix politics during rlhf, but having llm output in your training set is irrecoverable.

[–] Xylight@lemdro.id 1 points 1 day ago (1 children)

having llm output in your training set is irrecoverable

i used to think model collapse was an actual problem for LLMs as well, but it turns out that most popular models nowadays use intentionally synthetic data for things like reasoning traces and math. a lot of models (like gemini) also have subtle watermark patterns that let the trainers just filter out llm responses for factual data

[–] CombatWombat@feddit.online 1 points 1 day ago* (last edited 1 day ago)

Well, glad to hear LLM providers fixed that recently. I assume that means they'll stop taking my instance down now, yeah?