Meta has taught her artificial intelligence by illegally downloading books

Newly released documents show that Meta consciously used copyright sources to teach artificial intelligence models by unauthorized download.

Last month, Meta finally confessed to downloading a large and controversial data collection called Libgen from Torrent, which included tens of millions of robberies. Previously, details of the book collection had not been published, but Meta’s internal emails show that the company has at least 1.5 terabytes of data in several libraries, including at least 1.5 terabytes of Z-Library and Libgen with Toront. Downloaded.

Meta consciously used copyright books to teach artificial intelligence

Meta had also downloaded 1.5 terabytes of data from Libgen before. The publishers of these books have repeatedly accused meta of illegally using these resources to train artificial intelligence models, but Meta had always denied the allegations.

The authors have said in their complaint that the volume of meta -stealing resources is amazing. Of course, these sources are only 2.5 percent of the data that Meta has so far stolen. The authors had previously pushed Meta for more information, but Meta had rejected the authors’ request to review his data.

Meta has fully informed the copyright of these books. Nikolai Bashlikov, a researcher by researcher Meta, said in a message to his colleagues: “Download Torrent from the organizational laptop is not the right thing to do.” He also apparently expressed concern about using the Meta IP address to load the content stolen by Torrent.

Meta CEO Mark Zuckerberg had previously said that he had no interference in deciding on the use of stolen books to teach artificial intelligence, but disclosed messages show that he was also aware of the issue.