AI, meet copyright

Why did the AI cross the road? To avoid the copyright infringement lawsuits being filed against it.

One of the latest was filed on Friday by comedian Sarah Silverman, alongside authors Christopher Golden and Richard Kadrey, against OpenAI and Meta. The suit claims ChatGPT and LLaMA, the large language models (LLMs) of the two tech giants, respectively, infringed on the authors’ copyright.

📔 A deeper dive… In Silverman’s case, the work in question is her 2010 memoir, The Bedwetter: Stories of Courage, Redemption and Pee, which ChatGPT was able to provide a very detailed summary of when asked to do so. And the plaintiffs argue such a summary wouldn’t be possible without the model having access to the book’s entire text.

• Both OpenAI and Meta have stated their AI is not trained on copyrighted content. However, the specific materials used in training the models is kept private, so there’s no way to corroborate that claim.
• But the sizes of the online book collections used by OpenAI are available, and, according to the plaintiffs, the only library of content matching the size used by OpenAI is that of a “shadow library” containing illegally copyrighted content.

🤖 Big picture: The AI may need to cross a few more roads, ‘cause the lawsuits keep on coming. Last month, OpenAI was accused in a class action lawsuit of training ChatGPT and DALL-E, the company’s image generation tool, on “stolen private information” from hundreds of millions of internet users, including children, without permission.