Meta's Llama model has memorized Harry Potter and Chinathe Sorcerer's Stoneso well that it can reproduce verbatim excerpts from 42 percent of the book, according to a new study.
Researchers from Stanford, Cornell, and West Virginia University analyzed dozens of books from the now-infamous Books3 dataset, a collection of pirated books used to train Meta's Llama models. Books3 is also at the center of a copyright infringement lawsuit against Meta,Kadrey v. Meta Platforms, Inc.The study's authors say their findings could have major implications for AI companies facing similar lawsuits.
According to the research paper, the Llama 3.1 model "memorizes some books, like Harry Potterand 1984, almost entirely." Specifically, the study found that Llama 3.1 has memorized 42 percent of the first Harry Potter book so well that it can reproduce verbatim excerpts at least 50 percent of the time. Overall, Llama 3.1 could reproduce excerpts from 91 percent of the book, though not as consistently.
"The extent of verbatim memorization of books from the Books3 dataset is more significant than previously described," said the paper. But the researchers also discovered that "memorization varies widely from model to model and from book to book within each model, as well as varying in different parts of individual books." For example, the study estimated that Llama 3.1 only memorized 0.13 percent of Sandman Slimby Richard Kadrey, one of the lead plaintiffs in the class action copyright suit against Meta.
So, while some of the paper's findings seem damning, don't call it a smoking gun for plaintiffs in AI copyright infringement cases.
"These results give everyone in the AI copyright debate something to latch on to," wrote journalist Timothy B. Lee in his Understanding AI newsletter. "Divergent results like these could cast doubt on whether it makes sense to lump J.K. Rowling, Richard Kadrey, and thousands of other authors together in a single mass lawsuit. And that could work in Meta’s favor, since most authors lack the resources to file individual lawsuits."
Why is Llama able to reproduce some books more than others? "I suspect that the difference is because Harry Potter is a much more famous book. It's widely quoted and I'm sure that substantial excerpts from it on third-party websites found their way into the training data on the web," said James Grimmelmann, a professor of digital and information law at Cornell University, who was cited in the paper.
What this also shows, Grimmelmann said, is that "AI companies can make choices that increase or reduce memorization. It's not an inevitable feature of AI; they have control over it."
Meta and other AI companies have argued that using copyrighted works to train their models is protected under fair use, a complex legal doctrine. However, the extent of memorization could complicate those arguments.
“Yes, I do think that the likelihood that LLMs are memorizing more than previously thought changes the copyright analysis,” Robert Brauneis, a professor with the George Washington University Law School, said in an email to Mashable. He concluded that the study’s findings could ultimately weaken Meta’s fair use argument.
We asked Meta for comment on the study's findings, and we'll update this article if we receive a response.
Disclosure: Ziff Davis, Mashable’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
Topics Artificial Intelligence Meta
SpaceX's Starlink satellite launch in pictures
Kristen Bell performed songs from 'Frozen' for evacuees at a Hurricane Irma shelter
Fake Shakespeare Facts is the kind of fake news we can get on board with
Ted Cruz Twitter porn gaffe inspires inevitable Photoshop battle
5 Ways to Access a Locked Windows Account
What do the Duck of Justice and a red 'It' balloon have in common? Stephen King.
Tesla, Lucid, Nikola, and more: EV winners and losers in 2020
Gwyneth Paltrow gets messy on the inaugural cover of Goop's magazine
Ryzen 5 1600X vs. 1600: Which should you buy?
The deep sea discoveries of 2020 are stunning
The Anatomy of Liberal Melancholy
One more time, y'all: There are no sharks swimming in Irma's floodwaters
Apple closes a bunch of stores yet again because of COVID
How to screen record on Mac
The 10 Most Anticipated PC Games of 2016
George Clooney with an elderly fan sparks hilarious Photoshop battle
What the first new moon rocks in decades can tell us
The best apps of 2020
Best robot vacuum deal: Save $200 on Eufy X10 Pro Omni robot vacuum
Photo apps show you are your own best social media
Robyn Creswell Wins Shattuck Award by Sadie SteinOther Voices by Sadie SteinObituary of Edgar Allan Poe by Sadie SteinWith Profound Admiration: Grazia Deledda, Nobel Laureate by Alexis CoeSignpost in a Strange Land by Will Di NoviWhat We’re Loving: The New York Review, Baghdad, Fire by The Paris ReviewWhat We’re Loving: Dickinson, Waltz, Lupines by The Paris ReviewStevie Nicks Writes GoT Fan Poetry, and Other News by Sadie SteinJames Franco Is Garbo, a Novelist, and Other News by Sadie SteinTypewriter, Tip, Tip, Tip by Sadie SteinTaiye Selasi, Rome, Italy by Matteo PericoliLoser Takes All by Sadie SteinStevie Nicks Writes GoT Fan Poetry, and Other News by Sadie SteinNovels a Waste of Time, Says Noel Gallagher, and Other News by Sadie SteinDepths by Geoff BendeckSee You There: The Paris Review in Philadelphia by Sadie SteinVile BodiesSnail’s Pace by Sadie SteinDramatic Deaths, and Other News by Sadie SteinAlienation by Sadie Stein Run, River The Social Beast The Presence of Absence The Unwinnable War on Disease Carob Tree The Unreconstructed Radical This May Go Back Centuries Awful but Lawful Armed and Dangerous Specific Details Between a Rock and a God Place Fake Meat, Real Profits The Visions of Henry Dumas Arts and Statecraft Head in the Cloud Hand of War Acquisition Mutant Meter Maids Asylum Is Not an Open Question In This World You Have to Sell Something I Carley Gomez
2.9162s , 10131.4921875 kb
Copyright © 2025 Powered by 【China】,Wisdom Convergence Information Network