全球社交媒体巨头Meta近日承认在未经作者同意的情况下，使用受版权保护的书籍来训练其人工智能模型Llama 1和Llama 2。Meta使用了名为Books3的数据集，这是一个包含近20万本书籍的纯文本集合。然而，Meta辩称其使用受版权保护的作品来训练大模型不需要“同意、许可或付费”，并主张这种未经授权复制Books3中受版权保护的作品应被视为“合理使用”。此举引发了关于版权保护和人工智能训练材料使用的广泛讨论。
Title: Meta’s AI Training Controversy with Pirated Books
Keywords: Meta AI, Pirated Books, Training Controversy
News content: Meta, the social media giant, has recently admitted to using copyrighted books without authors’ consent to train its AI models Llama 1 and Llama 2. The company used a dataset called Books3, which contains nearly 200,000 books in pure text format. However, Meta argues that it does not require “consent, permission, or payment” to use copyrighted works to train its large models and claims that unauthorized copying of copyrighted works in Books3 should be considered “fair use.” This action has sparked widespread discussion about copyright protection and the use of training materials for AI.
【来源】https://www.techspot.com/news/101507-meta-admits-using-pirated-books-train-ai-but.html
Views: 1