If LLM-generated text increasingly fills the internet and becomes training data for future models, the system enters a feedback loop that collapses to a dull, stable state repeating the same thing forever—a vulnerability of language models that, like other feedback systems, becomes hard to model with Markov chains.
forecastpending
Speaker
Unknown SpeakerEvidence Quote
“it just says the same thing over and over and over again forever. The language models are vulnerable to this process”
Created: 6/18/2026, 1:59:31 PM
Your notes
Loading notes...