Meta filed its reply in support of its motion for summary judgment on fair use in AI training, including the torrenting of “pirated” copies of books datasets online. This brief is well-written. Will have more to say about it and Kadrey’s briefs, also well-written, in a subsequent post. Judge Chhabria will have much to ruminate over.
Excerpt:
Related Stories
One response to “Meta reply in support of fair use in AI training, including torrenting of “pirated” books datasets. Meta also opposes Kadrey book authors’ motion for summary judgment on infringement based on torrenting books.”
Not sure I like these Meta’s arguments. (1) There are circular reasonings here: Meta’s AI model, as Meta claimed, is “transformative”, so their copying or pirating of Plaintiffs’ books is fair. Meta’s copying of books is not infringing because they are training an AI that’s “transformative”. While this circular reasonings doesn’t remove the possibility of Meta’s AI being “fair use”, it’s really a weak argument. (2) Meta didn’t defend themselves on the “non-removal” of the CMI of books from Project Gutenberg by stating any reason why they did not remove it; they didn’t try to refuse Plaintiffs’ supposed findings by stating those findings are false. This can weaken Meta’s arguments about removing CMI “improve[s] model performance”.