AI companies are buying tons of old books because they can use them for model training, knowing they’re not filled with AI slop.
They’re not buying a whole bunch of copies of the same book (removing the book from circulation) — they’re buying all the unique books available to OCR and include in training data, because that falls under fair use.
If our government wanted to train its own AI model it already owns all the books and could just digitize the books we already have in the library. Then the AI could be free just like the library lol
Seems like an obvious solution if the government really wants to go all-in on AI: just make the LoC a training set; they could even sell access or license access if they wanted.
AI companies are buying tons of old books because they can use them for model training, knowing they’re not filled with AI slop.
They’re not buying a whole bunch of copies of the same book (removing the book from circulation) — they’re buying all the unique books available to OCR and include in training data, because that falls under fair use.
Fair use according to them. The court cases are ongoing. Personally I don’t see how them ingesting and commercializing is fair use.
If our government wanted to train its own AI model it already owns all the books and could just digitize the books we already have in the library. Then the AI could be free just like the library lol
Seems like an obvious solution if the government really wants to go all-in on AI: just make the LoC a training set; they could even sell access or license access if they wanted.