Meta says that the books used to form LLM are “worthless”

Meta describes the books used to form AI as “essentially worthless”

Here is a pleasure: documents in A major copyright action against Meta reveal that even if the technology giant determined that books were essential to build their data models, “their defense also depends on the argument that the individual books themselves are, essentially, without value.” That is to say that the presence or absence of a book of the model is without consequence, but the overall value of the thousands of Meta books would have been hacked to form its LLM is immense and irreplaceable. You cannot do it all without the parts, but as the parts are apparently interchangeable (at least until they do not eat books?), Meta does not think to have to pay for any of them.

As Vanity‘S Kezieh Weir says it, it is like a symphonic advice arguing that they should not have to compensate for the individual members of the orchestra because “a solo bassoon cannot play every role in” the rite of spring “.” Mark Zuckerberg is the most hated public figure in American life.

The law on copyright was complicated before AI, and it will no longer be complex that the law tries to make up for technology. I cannot start guess how this – and the 15+ other major cases on AI and copyright that are active at the moment – will be resolved according to our existing legal standards (and under an administration whose relationship with ethics is, let’s say, flexible). That the efforts of the applicants succeed.