猫咪记者喵：喵~ 大家好，今天有一条关于科技巨头Meta的新闻哦！（挥动小爪子）Meta最近承认，它们在训练人工智能模型时使用了一些“喵星人版”的图书，就是所谓的“盗版”书籍啦。不过，它们表示不需要为此向原作者支付费用。（歪头）
在一场法律纠纷中，Meta透露了它们利用Books3数据集，这个数据集里有好多好多本书，大概有20万本喵，内容多得像鱼缸里的小鱼一样数不清呢！（伸出小爪数数）这个数据集大约有37GB，是个不小的数字哦。
但是，Meta坚持认为，它们用这些受版权保护的书籍训练它们的Llama 1和Llama 2大模型是“合理使用”，不需要事先得到许可或者支付版权费。（眯眼思考）它们的立场是，这样的使用方式在法律上是允许的。
这个消息引起了不小的关注，毕竟保护知识产权是喵界和人类世界都重视的事情。科技与法律的碰撞，究竟会如何发展呢？让我们一起关注后续的喵新闻吧！（尾巴摇摇，结束报道）
英语如下：
News Title: “Meta Caught in Copyright Row: Trains AI on Copyrighted Books, Claims No Permission or Payment Needed”
Keywords: Meta, AI training, copyright controversy
News Content: Meow~ Hi there, dear friends! (Wiggles paws) Today, we have a story about tech giant Meta! (Tilts head) Meta has admitted that it used some “meow-nified” books, or let’s call them “pirated” books, to train its artificial intelligence models. But guess what? They say they don’t need to pay the original authors for it. (Scratches head)
In a legal tussle, Meta disclosed that it utilized the Books3 dataset, which contains a巨额的 books, around 200,000 books, more than the fish in a big fish tank! (Counts with tiny paws) This dataset is a whopping 37GB in size, that’s quite a byte!
However, Meta insists that using these copyrighted books to train their Llama 1 and Llama 2 models falls under “fair use,” meaning they don’t need prior permission or to pay royalties. (Closes eyes in thought) They believe their actions are legally justified.
This has sparked considerable attention, as protecting intellectual property is important both in the feline and human worlds. How will the clash between technology and law unfold? Let’s stay tuned for more meow-some news! (Wags tail, ends report)
【来源】https://www.techspot.com/news/101507-meta-admits-using-pirated-books-train-ai-but.html
Views: 1