**Meta承认利用“未经授权”的书籍训练人工智能，引发版权争议**
据TechSpot报道，科技巨头Meta在一项诉讼中透露，其在开发人工智能大模型Llama 1和Llama 2的过程中，使用了包含近20万本书的开源图书数据集Books3。这个数据集的总容量达到了近37GB，其中包括一些受版权保护的作品。然而，Meta声称在无需获得“同意、许可或付费”的情况下，使用这些版权作品进行大模型训练是合理的，认为这一行为应被视作“合理使用”。
这一声明立即引发了关于版权法和人工智能训练伦理的讨论。Meta的立场表明，该公司认为在人工智能研发中，使用此类数据可能是为了公众利益，符合“合理使用”的法律原则。然而，这一立场与传统的版权法规可能存在冲突，因为通常情况下，使用受版权保护的内容需要获得版权所有者的许可。
这一事件将Meta置于一个敏感的法律和道德十字路口，可能对未来的AI开发产生深远影响。随着技术的进步，如何在尊重知识产权和推动技术创新之间找到平衡，已经成为业界和法律界亟待解决的问题。此次诉讼的结果可能会为类似的案例设定先例，对整个科技行业产生广泛影响。
英语如下：
**News Title:** “Meta Acknowledges Using Unlicensed Books to Train AI, Refuses to Pay Royalties: Fair Use or Rights Infringement?”
**Keywords:** Meta, Artificial Intelligence, Copyright Dispute
**News Content:**
Tech giant Meta has admitted to utilizing an “unauthorized” collection of books to train its artificial intelligence models Llama 1 and Llama 2, sparking a copyright controversy, according to TechSpot. The dataset in question, Books3, comprises nearly 200,000 books, amounting to almost 37GB, some of which are protected by copyright. However, Meta claims that using these copyrighted works for large-scale model training without “consent, permission, or payment” falls under the umbrella of “fair use,” asserting that this practice should be considered acceptable under the legal principle.
This statement has immediately ignited discussions around copyright law and the ethics of AI training. Meta’s stance implies that the company believes using such data in AI development is for the public good, aligning with the “fair use” doctrine. Nevertheless, this position may clash with traditional copyright regulations, which generally necessitate obtaining permission from copyright holders for the use of protected content.
The incident places Meta at a delicate juncture of legal and ethical implications, with potential far-reaching effects on future AI development. As technology advances, striking a balance between respecting intellectual property rights and fostering innovation has become an urgent issue for both the tech industry and legal community. The outcome of this lawsuit could set a precedent for similar cases, exerting wide-ranging influence across the technology sector.
【来源】https://www.techspot.com/news/101507-meta-admits-using-pirated-books-train-ai-but.html
Views: 1