Meta's official story is that nobody at the company ever decided to pirate boo… · @harden_369741 · TwKit

@harden_369741

Meta's official story is that nobody at the company ever decided to pirate books to train Llama. A federal judge just ordered them to hand over the names anyway. I buy my ebooks one at a time like everyone else, so the idea that flagship models were trained on torrented books while no executive signed off on it genuinely rattles me. The discovery order from Magistrate Judge Thomas Hixon says Meta must identify every person who "made, authorized, or approved" the torrenting from Anna's Archive and LibGen. Meta's defense: there was no formal decision, just "collaborative" talks, so there is nobody to name. The judge did not buy it. My takeaway: "no formal decision" is the new "the intern did it." Informal process is exactly how accountability evaporates, and that is why the court wants names, not org charts. If your training data came from a torrent, at what point does "we all just talked about it" stop being an excuse? #Copyright #AI

曝光 25 · 评论 1 · 点赞 0 · 书签 0