FWIW It's not clear cut if AI generated data feeding back into further training reduces accuracy, or is generally harmful.
Multiple papers have shown that generated images by high quality diffusion models with a proportion of real images in mix (30-50%) improves the adversarial robustness of the models. Similiar things might apply to language modeling.
10239 post score 540 comment score https://old.lemmy.world/u/awesome_person@lemm.ee
Why does Lemmy make it look harder than it is? It's not a massive load compared to what modern servers and applications are designed to handle.
I couldn't sign up on beehaw and lemmy.ml after multiple tries. It feels worse than a simple centralised platform one can build in a month.
Is there alternative to reddit for people like me who don't need this kind of decentralisation (Lemmy feels like centralisation, just multiple number of it, if any instance can cut off like this.) but likes the (text heavy)interface of Lemmy?