(C) Daily Kos
This story was originally published by Daily Kos and is unaltered.
. . . . . . . . . .
AI "Model Collapse" looms as tech companies try to train systems using an internet full of AI slop [1]
['This Content Is Not Subject To Review Daily Kos Staff Prior To Publication.']
Date: 2025-06-19
Excellent link, please click me.
The TL:DR summary is that companies trying to train “AI” systems (sarcastic quote marks are sarcastic) using information produced by humans are running out of information produced by humans and resorted to allowing their systems to troll the Internet for info to fill gaps in their data.
I.e. “just google it.”
Since you’re a DKos reader, and therefore presumably intelligent, I’ll wager you already grasp the core of the issue here.
Problem is, given:
1) ‘link rot’ which has led to a lot of older [pre-AI] net content being no longer available (“the internet never forgets” is a myth) and
2) the sheer volume of dreck being spewed out by bots and other nonhuman content makers,
the Internet is (as we all know) is already neck-deep in AI-generated garbage that clogs up all the results on any search method, so the new version AIs are being trained on what’s increasingly the product of earlier, even-less-good AIs, for a recursive cycle of suck.
For me, the Jenga moment of this article was:
A third option, Vaughn-Nichols predicts, appears to already be in motion. "We're going to invest more and more in AI, right up to the point that model collapse hits hard and AI answers are so bad even a brain-dead CEO can't ignore it," he wrote.
Yeah that sounds like a solid prediction, given how US tech and financial industries have historically behaved ever since the …………… um pretty much since….. ever? FWIW I seriously wrote and discarded about a dozen ideas there but kept thinking of an older, previous Charlie Foxtrot, dating back all the way to the 1720 South Seas bubble, at which point I gave up. There’s some data to train a model on, how ‘bout that?
The good news is, the probably means no Skynet.
The bad news is, Idiocracy apparently applies to AIs too.
[END]
---
[1] Url:
https://www.dailykos.com/stories/2025/6/19/2329024/-AI-Model-Collapse-looms-as-tech-companies-try-to-train-systems-using-an-internet-full-of-AI-slop?pm_campaign=front_page&pm_source=more_community&pm_medium=web
Published and (C) by Daily Kos
Content appears here under this condition or license: Site content may be used for any purpose without permission unless otherwise specified.
via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/dailykos/
