Honestly, I actually believe the books2 corpus controversy is true. Where else, other than library genesis, would they come across 500B tokens of books? Conside

Honestly, I actually believe the books2 corpus controversy is true. Where else, other than library genesis, would they come across 500B tokens of books? Considering books1 is 50B and they describe that as open domain books, what's left would seem to me to be books that they don't have the rights to use or right to have digitally. I'm not bemoaning them for it. I'd do the same thing because I don't really give af about copyright laws. It also seems obvious by the way each were treated in the training process. They pretty much say that while books1 corpus was trained on several times, they used books2 far less. That's exactly what I would say to cover my ass so that when I'm caught, I can tell them, "Look, though, we barely even used it. In fact, the paper overstates our use. As far as remember, we only used the table of contents. And not even the whole table of contents. Just a fraction of it. Are you gonna arrest me for using a fraction of the table of contents?"