AI translated articles swapped sources or added unsourced sentences with no explanation, while others added paragraphs sourced from completely unrelated material.
Ironically, the actual human contributions to Wikipedia play a significant role in training LLMs. And since you can’t train LLMs on LLM generated text without recursion errors, the more Wikipedia is contaminated by AI text, the less useful Wikipedia becomes for training purposes.
I’m torn between thinking “keep poisoning your own fucking well, assholes” and “look at all this free labor Wikipedia editors are doing to help AI developers”.
Ironically, the actual human contributions to Wikipedia play a significant role in training LLMs. And since you can’t train LLMs on LLM generated text without recursion errors, the more Wikipedia is contaminated by AI text, the less useful Wikipedia becomes for training purposes.
I’m torn between thinking “keep poisoning your own fucking well, assholes” and “look at all this free labor Wikipedia editors are doing to help AI developers”.
They won‘t poison old books and such, the internet didn‘t keep any promises anyway.