The times, as Dylan sang, are a changing.
— — —
The New York Times sued OpenAI hoping for billions of dollars in damages. You see, Gray Lady believes it is among the largest sources of proprietary information for OpenAI and Microsoft’s AI products. Let’s pretend they’re right.
The Times said they spend hundreds of millions per year to create this media. They say OpenAI took that media without asking and didn’t pay for it. Not nice. And, importantly, those are statements of fact.
Tech companies counter with argument with, “Fair use.” People can take pieces of media to comment on, criticize, or parody. I’m not a lawyer. So, I lifted that nearly verbatim from the good folks at Stanford Law. They used the word, “transformative” as a way to think about whether or not the use is fair.
A lot of people will focus on how to define this, “IS.” Did OpenAI transform the New York Times’ words? The Times, of course, says, “No.” They continue to say, “The fair use argument shouldn’t apply, because the AI tools can serve up, almost verbatim, large chunks of text from Times news articles.”
“Can,” is interesting here. As in “…can serve…” OpenAI *can* serve up copy verbatim. But, for the most part, they don’t. ChatGPT is an application of OpenAI. You can ask it to write things. Like, a wedding toast or a book report. Will it serve up verbatim text from the New York Times? Not likely. If you asked it to describe quantum mechanics or New York election history, OpenAI’s answer comes from multiple sources. Once in a while, to a very specific question, you might get a specific verbatim answer. I wrote about it the other day. Mostly, you’re going to get a synthesized answer from many sources glommed together in a particular style.
I don’t think Fair Use applies here nearly as widely as The Times would like it to. But, again, I’m not a lawyer.
Let’s say The Times wins its argument. The case pivots to damage. And, this is where I think their case unravels.
“Their [OpenAI’s] AI tools divert traffic that would otherwise go to the Times’ web properties, depriving the company of advertising, licensing and subscription revenue.”
With the NYT content, OpenAI knows the PAST. The NYT’s reporters write about the NOW. The use cases are different. The traffic isn't the same. The NYT may have a case, but this isn't the one.
This feels — a lot — like when the USFL sued the NFL. They had a solid anti-trust case. Sued for $1.69B (1986 dollars — the good stuff). And won. The jury awarded them ONE DOLLAR.
And here’s why, “It's as blatant as the nose on your face [that the NFL is a monopoly]. . . . The thing is, how was anyone hurt by them?” In this case the jury determined — after 31 hours of stormy deliberation — that the NFL monopoly had not significantly injured the USFL.
Ironically, I got these USFL deets from Sports Illustrated — when it was not written by AI.
I’m going to guess this story was in the New York Times too. They know this. I think this suit is meant to get leverage. They’ll settle.
Addendum…
The proprietary, ever-fresh data sets that Meta (Facebook, Instagram, WhatsApp), Google (YouTube, Gmail; search inputs – not snippets), Microsoft (Word, Outlook, PowerPoint), Twitter/X, Apple, and Snap – among others have — are worth BIG money.
OpenAI was very smart to pair up with a company that has lots of users sending it training data. Without Microsoft, they would be on the outside looking in.