I doubt it. From what I understand the info that's out there already is more that enough to bootstrap useful human level intelligence. Anything people make in

I doubt it. From what I understand the info that's out there already is more that enough to bootstrap useful human level intelligence. Anything people make in the near future isn't going to be that radically different. Sure there will be excellent essays and books and organizational data and slide, etc... But including it in the 3.6 Trillion Tokens would be a tear in the ocean. Unless you think someone is going to create such a monumentally radical set of non-intuitive token relationships that outstrip the possible use of the rest? Maybe? TL;DR - It's the scale. Wait, sorry, that doesn't answer your actual question. For some reason I thought you were asking "Will the siloing of new data make for crappier LLMs?"