Was about to share this one. Dupe alert caught it.

Putting ethics questions aside, I actually think this is a great consequence of this kind of training: knowledge that has been buried and forgotten in books for decades now finally gets to live and flourish. It'd be great if during the same procedure, the proper referencing of each book emerges from adequate training. Citations are still infected with hallucinations when using ChatGPT.

south_korea_ln

It depends on if they are public domain or not I would imagine. I wondered about how much modern text you need to learn the new concepts

I don't think they can do anything with it except train their own models on it?

optimism

Is this that bad? Now there are digital copies? Perhaps release it to project gutenberg