Meta purportedly trained its AI on more than 80TB of pirated content and then open-sourced Llama for the greater good
Posted by:
James Thompson
Meta, the parent company of Facebook, is reportedly embroiled in controversy over its AI training methods, as revealed in court documents. The company is facing a class-action lawsuit accusing it of copyright infringement and unfair competition in training its AI model, Llama. Meta allegedly downloaded a significant amount of pirated books, around 82TB, from shadow libraries like Anna’s Archive and LibGen for this purpose.
Internal discussions within Meta showed that some employees raised ethical concerns as early as 2022 regarding the use of pirated materials. Despite objections, Meta continued its AI training activities and took measures to avoid detection, including discussions on evading the tracking of downloads using corporate IP addresses.
Meta’s CEO, Mark Zuckerberg, reportedly pushed for the implementation of AI technology within the company in a meeting held in January 2023, despite internal objections. This controversy is not unique to Meta, as other tech giants like OpenAI and Nvidia are also facing legal challenges over the use of copyrighted materials in their AI training processes.
The news sheds light on the ethical and legal dilemmas surrounding AI training practices, with concerns raised about the potential violation of copyright laws and ethical standards in the pursuit of technological advancements.
Recent Posts1
