This is essentially what the New York Times is reporting in its lawsuit against OpenAI. The Times (Mothra) is the lifeblood of liberal democracy in the fight against the Big Tech disruptors. OpenAI (Godzilla) has already looted the Times by using the article to train his ChatGPT without permission. The inclusion in the lawsuit of ChatGPT’s results, which are a near-word-for-word copy of the New York Times article, certainly makes OpenAI look very ugly.
OpenAI completely rejects this assumption. Surprisingly, until recently, he had been negotiating with the Times for rights fees. In a blog post written in the tone of a man who doesn’t know why his date left the restaurant, the company said that “discussions with the New York Times appeared to be progressing constructively.” …We believe the New York Times’ lawsuit is without merit. Still, we look forward to a constructive partnership with The New York Times and respect its long history. ”
There are several cases where copyright and generative AI intersect. Sarah Silverman vs. Meta, Getty Images vs. Stability AI –But this is the only thing that can tempt Don King out of retirement. These are culturally dominant apex companies with reputations for innovation and arrogance. But it’s also a story of the underdog. The Times is worth about $8 billion. OpenAI is valued at about $100 billion, but its largest shareholder, Microsoft, is worth 16 people, including Jeff Bezoses.
As a journalist who writes about artificial intelligence, all people in my profession want to talk about is NYT vs. OpenAI. And they all have the same question. “How will it end?”
Elon Musk famously declared:The most interesting outcome is the most likely.This is obviously insane, but it’s solid advice for writing columns. So let’s enjoy Musk’s razor and dig a little deeper into three possibilities: (1) Verdict. (2) Congressional action to clean up copyright law and avoid the need for a verdict. (3) Settlement.
The funniest outcome is clearly the victims of such heinous harm that the New York Times took the case all the way to the Supreme Court, where the justices wearing Tom Friedman’s mustaches unanimously ruled in favor of the Gray Lady. will prove that it is.
Now comes the interesting part. The court then agrees to the Times’ request that OpenAI destroy chatbot models and training data that use Times’ copyrighted material. This is known as algorithmic disgorgement. This is a legal concept to address illicit profits from algorithmic sleaze. Regular technology companies may be forced to remove infringing materials from their products at significant cost. But generative AI is no ordinary technology. It is created by a combination of hundreds of billions of parameters of information and language. That makes extracting the New York Times from ChatGPT as impossible as pulling eggs out of a cake. If subject to disgorgement, OpenAI could theoretically be forced to delete everything it used to train ChatGPT and start over, effectively putting the company out of business. Wordle for the next day: Pwned.
I don’t think this will happen. First of all, OpenAI has a decent fair use argument based on his ChatGPT’s ability to transform the Times article into something original. (OpenAI has dismissed the Times’ plagiarism claim as a “rare bug.”) Considerations of social benefit are also part of fair use. As a result, news organizations are often allowed to use copyrighted material, and poets are often able to use excerpts from poems. Socially, we decided we wanted to encourage them. You can debate whether using ChatGPT is a social good, but tens of millions of Americans are already using it. Disgorgement, or even an injunction, would not only harm OpenAI, but also those people. Finally, the legal process could take years, during which time ChatGPT could become even more entrenched. Think about how Uber showed up in cities, pushed aside taxis, and never asked for forgiveness. You don’t have to be just to win in court. It just has to become inevitable.
Timely Congressional action also seems unlikely. If you require further clarification, we will refer you to the Post’s publishing department. Our staff will then ask you a few questions about your reason for subscribing.
This would lead to a reconciliation, but this sounds anti-climactic. But given the current positions of both parties, the negotiations are likely to see more drama than a slow-motion legal battle.
AG Sulzberger, chairman of the New York Times Company and publisher of the New York Times, believes he had billions of dollars in intellectual property stolen, according to people familiar with his thinking. He says he believes it. He knows that OpenAI has already ingested the Times’ content and cannot retrieve it. Even if he makes OpenAI the godfather offer, a one-time payment is a non-starter because today’s windfall doesn’t guarantee that AI won’t swallow up his family business one day. The New York Times, which was nearly wiped out by the first wave of AI-powered social media algorithms, sees an opportunity for tech companies to earn huge licensing fees in long-term contracts, or billions of dollars in recurring revenue. ing. And when it can’t get a number, the company has a proud history of going to court to uphold its principles.
OpenAI, on the other hand, is at the top of its “let’s burn the cash on our way to market domination” phase. Everyone is dazzled by the possibilities, so it’s the perfect moment to shell out money and throw a one-time sum at the New York Times. However, there are no large recurring payments. The last thing OpenAI wants is to become a royalty administrator and constantly argue percentages with every small publisher who can claim to have attended ChatGPT training. Not only is it cumbersome and a waste of future profits, it’s also unglamorous by Silicon Valley standards.
Since filing the lawsuit, the Times has done its best to enjoy its influence in the protection of silence. (To be fair, the company continues to keep its ethical walls high, too. None of the Times reporters I spoke with had any idea what management was thinking about the incident.) (That was very frustrating.) OpenAI, on the other hand, was like the Jay Gatsby of the White House. Dinner with correspondents, buy drinks, and close deals with all other news organizations whenever possible. Although these transactions are worth less than a few million dollars, they represent great generosity to the Fourth Estate. It’s the perfect way to show the public and the courts that OpenAI isn’t stubborn.
The potential for an eventual settlement, perhaps within the next few months, is that both sides are at a small existential risk. The Times is one of the best economic articles in the American journalism wilderness, but AI has brought huge new unknowns. It now seems prudent to ensure its sustainability. OpenAI has already had one near-death experience, and while it is hard to imagine a disgorgement verdict, it was forced to admit that ChatGPT is addicted to copyright. “It would be impossible to train today’s leading AI models without using copyrighted material,” the company wrote in a submission to the House of Lords.
This last part is what makes me think OpenAI is likely to bend. Now is the time for the company to decide on the price of its drug of choice. When major record labels battled with Spotify, they ultimately reached an agreement to license music to the platform while acquiring an 18 percent stake in the company. Spotify didn’t want that (and OpenAI won’t go that high either), but music services just can’t live without music. Neither party is excited about the relationship, but this is a form of mutually assured relationship maintenance. Would it work to allocate a 5% stake in OpenAI pro rata to the largest news organizations and book publishers, rebalancing once a year to avoid excessive accounting?
In the movie, Godzilla kills Mothra with his fiery atomic breath, and Mothra’s children take revenge by wrapping Godzilla in silk and causing him to limp like a lizard and fall into the ocean to die. These results can be quite thrilling. But there’s so much at stake, and with apologies to Mr. Musk, the outcome is most likely the least interesting.
