Post #1722610
2026-04-27 12:13 UTC
RE: https://universeodon.com/@devsimsek/116472552957401849
I'm less interested in the mathematical prove and rather in the concept of 'model collapse' and that it forgets what reality is like. That sounds wild. LLM is k-holeing? I mean, AGI is based on the wrong assumption that intelligence can scale, whatever that means. Intelligence is at best described as an assemblage of affordances and enactivism, and never not as a unit of measurement. That already creates a fallacy where people attempt to measure and apply it economically. Super unhealthy, let alone IQs rooting in eugenics and racism.
Replies (1)
-
@a@pdx.social 2026-04-27 14:06
@thgie What’s frustrating is that while this “proof” might be new, model collapse has been known—and observed!—for years. But even when it’s observed, sycophants and the very gullible say these things are “creating new languages” or whatever, as though it’s a step *towards* AGI… no, they’re just breaking, failing to do the job for which they were built! The term “model collapse” did help with the meter for my Taylor Swift rewrite: https://pdx.social/@a/113400143539613625