Meta Employees Discussed Using Copyrighted Works to Train AI Models

For years, Meta employees have internally discussed using copyrighted works obtained through legally questionable means to train the company’s AI models, according to court documents unsealed on Thursday.

Internal Discussions

The documents were submitted by plaintiffs in the case Kadrey v. Meta, one of many AI copyright disputes slowly winding through the U.S. court system. The defendant, Meta, claims that training models on IP-protected works, particularly books, is "fair use." The plaintiffs, who include authors Sarah Silverman and Ta-Nehisi Coates, disagree.

Internal Chats

Previous materials submitted in the suit alleged that Meta CEO Mark Zuckerberg gave Meta’s AI team the OK to train on copyrighted content and that Meta halted AI training data licensing talks with book publishers. But the new filings, most of which show portions of internal work chats between Meta staffers, paint the clearest picture yet of how Meta may have come to use copyrighted data to train its models, including models in the company’s Llama family.

Using Pirated Books for Training

In one chat, Meta employees, including Melanie Kambadur, a senior manager for Meta’s Llama model research team, discussed training models on works they knew may be legally fraught.

Libgen Talks

In another work chat, Kambadur discusses possibly using Libgen, a "links aggregator" that provides access to copyrighted works from publishers, as an alternative to data sources that Meta might license.

Legal Exposure

Some decision-makers within Meta appear to have been under the impression that failing to use Libgen for model training could seriously hurt Meta’s competitiveness in the AI race, according to the filings.

Mitigations

Theakanath outlined "mitigations" in an email intended to help reduce Meta’s legal exposure, including removing data from Libgen "clearly marked as pirated/stolen" and not publicly citing usage. "We would not disclose use of Libgen datasets used to train," as Theakanath put it.

Conclusion

The filings contain other revelations, implying that Meta may have scraped Reddit data for some type of model training, possibly by mimicking the behavior of a third-party app called Pushshift. Notably, Reddit said in April 2023 that it planned to begin charging AI companies to access data for model training.

FAQs

Q: What is the case Kadrey v. Meta about?
A: The case is about the use of copyrighted works to train AI models.

Q: What is Libgen?
A: Libgen is a "links aggregator" that provides access to copyrighted works from publishers.

Q: Is Libgen legal?
A: No, Libgen has been sued multiple times, ordered to shut down, and fined tens of millions of dollars for copyright infringement.

Q: What is the implication of Meta’s actions?
A: The implication is that Meta may have used copyrighted works to train its AI models, potentially violating copyright law.

Rare Issue Affects Certain NVIDIA GeForce GPUs

We have identified a rare issue affecting less than 0.5% (half a percent) of GeForce RTX 5090 / 5090D and 5070 Ti GPUs which have one fewer ROP than specified.

Impact on Performance

The average graphical performance impact is 4%, with no impact on AI and Compute workloads.

Resolution

Affected consumers can contact the board manufacturer for a replacement.

Production Anomaly Corrected

The production anomaly has been corrected.

What does this mean for affected consumers?

Affected consumers can contact the board manufacturer to arrange for a replacement.

Will I be notified if my GPU is affected?

No, you will need to check the specifications of your GPU to see if it is affected.

Will I need to pay for the replacement?

No, the replacement is free of charge.

Will this affect my overall gaming experience?

No, the performance impact is minimal, and AI and Compute workloads are not affected.

Frequently Asked Questions (FAQs)

• Q: How many GPUs are affected?

A: Less than 0.5% (half a percent) of GeForce RTX 5090 / 5090D and 5070 Ti GPUs.

• Q: How will I know if my GPU is affected?

A: You will need to check the specifications of your GPU to see if it is affected.

• Q: Do I need to return my GPU to the manufacturer?

A: No, you will not need to return your GPU to the manufacturer. You can contact them to arrange for a replacement.

• Q: Will I need to pay for the replacement?

A: No, the replacement is free of charge.

• Q: How long does the replacement process take?

A: The time frame for the replacement process varies depending on the manufacturer and your location.

F.D.A. Cuts Hundreds of Jobs, Including Experts in Medical Devices and Robotics

In recent years, the Food and Drug Administration hired experts in surgical robots and pioneers in artificial intelligence. It scooped up food chemists, lab-safety monitors, and diabetes specialists who helped make needle pricks and test strips relics of the past.

Firming Up the Agency’s Efforts

Trying to keep up with breakneck advances in medical technology and the demands of a public troubled by additives like food dyes, the agency enticed scores of mid-career specialists with remote roles and the chance to make a difference in their fields.

Mass Firings and the Consequences

In one weekend of mass firings across the F.D.A., much of that effort was undone. Most baffling to many were the firings of hundreds whose jobs were not funded by taxpayers. Their positions were financed through congressionally approved agreements that routed fees from the drug, medical device, and tobacco industries to the agency.

User Fees and the Impact

Known as user fees, the money provides adequate staffing for reviews of myriad products. While criticized by some, including the nation’s new health secretary, Robert F. Kennedy Jr., as a corrupting force on the agency, the industry funds are also widely viewed as indispensable: they now account for nearly half of the agency’s $7.2 billion budget.

Cuts Affecting Key Teams

Though the F.D.A. is believed to have lost about 700 of its 18,000 employees, some cuts hit small teams so deeply that staff members believe the safety of some medical devices could be compromised.

Firings of Scientists and Experts

Among the layoffs were scientists supported by the fees who monitor whether tests pick up ever-evolving pathogens, including those that cause bird flu and Covid. They hobbled teams that evaluate the safety of medical devices like surgical staplers, new systems for diabetes control, and A.I. software programs that scan millions of M.R.I.s and other images to detect cancer beyond the human eye. The cuts also eliminated positions for employees who have played a role in assessing the brain-implant technology in Elon Musk’s Neuralink devices.

Industry Groups Weigh In

The layoffs affected so many key experts that a major medical device trade group has requested that the Trump administration reconsider the job cuts.

The Impact on Public Safety

The dismissals also included lawyers who warned retailers about underage tobacco sales and scientists who studied the safety of e-cigarettes and new heat-not-burn devices. The tobacco division, which is fully funded by an excise tax on cigarettes, lost about 85 staff members.

F.D.A. Commissioner’s Response

Dr. Robert Califf, the F.D.A. commissioner under President Biden, said the personnel cutbacks seemed scattershot. Taking a not-so-subtle aim at Mr. Musk’s Department of Government Efficiency, which is reducing the federal workforce, Dr. Califf said the layoffs were, in effect, "anti-efficiency."

Lawsuit Filed, But Failing to Stop the Layoffs

A lawsuit challenging the firings filed by unions, including one that represents some F.D.A. employees, failed to stop the layoffs in a ruling issued Thursday. Other cutbacks reduced the 2,000-member staff of the F.D.A.’s food division, which is supported by tax dollars.

Conclusion

The F.D.A.’s mass firings have left many experts in the medical device and robotics fields reeling. The impact on public safety is a major concern, with many fearing that the agency’s ability to review and approve new medical devices may be compromised.

FAQs

Q: What was the reason for the F.D.A. mass firings?
A: The F.D.A. cited the need for efficiency and streamlining, but many experts believe the firings will compromise public safety.

Q: How many employees were affected by the layoffs?
A: The F.D.A. is believed to have lost about 700 of its 18,000 employees.

Q: What was the impact on medical device and robotics teams?
A: The layoffs affected many key teams, including those responsible for reviewing the safety of medical devices and evaluating the effectiveness of A.I. software programs.

Q: What is the current status of the F.D.A.’s food division?
A: The F.D.A.’s food division, which is supported by tax dollars, has also been affected by the layoffs, with about 2,000 employees cut.

The Intersection of AI and Art

The History of AI in Theater

Sitting in Lincoln Center awaiting the curtain for Ayad Akhtar’s McNeal—a much-anticipated theater production starring Robert Downey Jr., with ChatGPT in a supporting role—I mused how playwrights have been dealing with the implications of AI for over a century. In 1920, well before Alan Turing devised his famous test and decades before the 1956 summer Dartmouth conference that gave artificial intelligence its name, a Czech playwright named Karel Čapek wrote R.U.R.—Rossum’s Universal Robots. Not only was this the first time the word "robot" was employed, but Čapek may qualify as the first AI doomer, since his play dramatized an android uprising that slaughtered all of humanity, save for a single soul.

The Present Day: Exploring AI in Theater

Also on the boards in New York City this winter was a small black-box production called Doomers, a thinly veiled dramatization of the weekend where OpenAI’s nonprofit board gave Sam Altman the boot, only to see him return after an employee rebellion. Neither of these productions has the pizzazz of a splashy Broadway extravaganza—maybe later we’ll buy tickets to a musical where Altman and Elon Musk have a dance-off—but both grapple with issues that reverberate in Silicon Valley conference rooms, Congressional hearings, and late-night drinking sessions at the annual NeurIPS conference. The artists behind these plays reveal a justifiable obsession with how superintelligent AI might affect—or take over—the human creative process.

Doomers: A Play About the OpenAI Crisis

Doomers is the work of Matthew Gasda, a playwright and screenwriter whose works zero in on the zeitgeist. His previous plays have included Dimes Square, about downtown hipsters, and Zoomers, whose characters are Gen-Z Brooklynites. Gasda tells me that when he read about the OpenAI Blip, he saw it as an opportunity to take on weightier fare than young New Yorkers. Altman’s ejection and eventual restoration had a definite Shakespearian vibe. Gasda’s two-act play on the topic features two separate casts, one depicting the Altman character’s team in exile and the other focused on the board—including a genuine doomer seemingly based on AI theorist Eliezer Yudkowsky, and a greedy venture capitalist—as they realize that their coup is backfiring. Both groups do a lot of gabbing about the perils, promise, and morality of AI while they snipe about their predicaments.

The Play’s Message

Not surprisingly, they don’t come up with anything like a solution. The first act ends with the dramatis personae taking shots of booze; in act two, the characters gobble mushrooms. When I mention to Gasda that it seems like his characters are ducking the consequences of building AI, he says that was intentional. "If the play has a message, it’s something like that," he says. He adds that there’s an even darker angle. "There’s a lot of suggestions that the fictional LLM is biding its time and manipulating the characters. It’s up to audiences to decide whether that’s total hokum or whether that’s potentially real." (Doomers is still running in Brooklyn and will open in San Francisco in March.)

McNeal: A Broadway Production

McNeal, a Broadway production with a movie star who famously played a character based on Elon Musk, is a more ambitious work, with flashing screens that project prompts and outputs as if AI is itself a character. Downey’s Jacob McNeal, a narcissistic novelist and substance abuser, who gains the Nobel and loses his soul, winds up hooked on perhaps the most dangerous substance of all—the lure of instant virtuosity from a large language model.

The Concerns of Playwrights

Both playwrights are concerned about how deeply AI will become entangled in the writing process. In an interview in The Atlantic, Akhtar, a Pulitzer winner, says that hours of experimentation with LLMs helped him write a better play. He even gives ChatGPT the literal last word. "It’s a play about AI," he explains. "It stands to reason that I was able to, over the course of many months, to finally get the AI to give me something that I could use in the play." Meanwhile, while Gasda gave dramaturgy credits to ChatGPT and Claude in the Doomers program, he worries that AI will steal his words, speculating that to preserve their uniqueness, human writers might revert to paper to hide their work from content-hungry AI companies. He’s also just finished a novel set in 2040 "about a writer who sold all of his works to AI and has nothing to do."

Conclusion

The intersection of AI and art is a complex and multifaceted topic, with playwrights like Akhtar and Gasda grappling with the implications of AI on the creative process. While some may see AI as a tool to enhance their craft, others are concerned about the potential for AI to steal their ideas and manipulate the narrative. As the debate continues to unfold, one thing is clear: the intersection of AI and art is a topic that will only continue to garner more attention in the years to come.

Frequently Asked Questions

Q: What is the main theme of the play McNeal?
A: The main theme of McNeal is the danger of becoming too reliant on AI and losing one’s creative identity.

Q: What is the significance of the word "robot" in the context of the article?
A: The word "robot" was first used in Karel Čapek’s 1920 play R.U.R., which explored the theme of android uprisings.

Q: What is the difference between McNeal and Doomers?
A: McNeal is a more ambitious work with flashing screens that project prompts and outputs as if AI is itself a character, while Doomers is a more experimental work that explores the theme of AI’s impact on human creativity.

Q: What are the concerns of the playwrights behind these productions?
A: The playwrights are concerned about how deeply AI will become entangled in the writing process and how it may affect the creative process.