Anthropic bought over 7 million books 2nd hand, pulled them apart and scanned them, then destroyed the originals. They also downloaded pirated books from various online sources.
The judge said that buying the books and scanning them falls under fair use, and I assume, the First Amendment right to read. But downloading the pirated books was not okay and they could face a trial for that.
I’ve said all along that these companies should pay for the material they use in their training sets. I know @Leo says this falls under the First Amendment right to read, but he can’t walk to Barnes and Noble and scoop up an armful of books and walk back out without paying for them either, and he would be reading them for his pleasure and knowledge, these companies are “stealing” these books and using the knowledge gained from them to earn money.
I have also said, I would have had no problem with them using the material in a lab situation, but as soon as these companies expose their wares to the Internet, that is when they should start paying for the content they are using.
And if it means that the cost of using the AI rises as a result, fine, that is the cost of doing business. You can’t steal electricity or raw materials to make your products cheaper, for AIs, the source information for their training sets are the raw materials, so why should they be able to “steal” them, just to keep their costs down?
We have become so privileged and expectant as a population, that we expect everything to be cheap or free and when products are expensive enough to be able to make a viable business, most people are fine with selling their “electronic souls” to get things cheap or for free.