I’ve got 99 problems with A.I., but intellectual property ain’t one.
Media and entertainment industries have lately been consumed with questions about how content generated by artificial intelligence systems should be considered under intellectual property law. Last week a federal judge ruled against an attempt to copyright art produced by a machine. In July another federal judge suggested in a hearing that he would most likely dismiss a copyright infringement lawsuit brought by artists against several artificial intelligence art generators. How A.I. might alter the economics of the movie and TV business has become one of the primary issues in the strike by writers and actors in Hollywood. And major news companies — including The Times — are weighing steps to guard the intellectual property that flows from their journalism.
In the face of all the possible actions against A.I. and its makers, I’d suggest caution. I’ve been thinking a lot about whether musicians, painters, photographers, writers, filmmakers and other producers of creative work — including, on good days, myself — should fear that machines might damage their livelihoods. After extensive research and consultation with experts, I’ve arrived at a carefully considered, nuanced position: meh.
Controversies over A.I. are going to put a lot of copyright lawyers’ kids through college. But the more I use ChatGPT, Midjourney and other A.I. tools, the more I suspect that many of the intellectual property questions they prompt will prove less significant than we sometimes assume. That’s because computers by themselves cannot yet and might never be able to produce truly groundbreaking creative work.
Indeed, I’d bet that artists and creative industries will ultimately find A.I. to be more of a boon than a competitor. In a recent assessment of A.I.-produced comedy, Jason Zinoman, The Times’s comedy critic, suggested that A.I. comedians might improve human comedy: “Competition from increasingly clever computer programs will force artists to not only rely more on intuition than imitation, but also to think harder about what makes them, and their work, distinctly human.”
I think he’s right — not just about comedy but also about many other creative fields. What accounts for my sunny stance? History offers one clue: Technologies that made art easier to produce have rarely ended up stifling human creativity. Electronic synthesizers didn’t eliminate the need for people who play musical instruments. Auto-Tune didn’t make singing on pitch obsolete. Photography didn’t kill painting, and its digitization didn’t obviate the need for professional photographers.
Then there’s the content I’ve seen A.I. produce: Unless it’s been heavily worked over by human beings, a lot of the music, images, jokes and stories that A.I. has given us so far have felt more like great mimicry than great art. Sure, it’s impressive that ChatGPT can write a pop song in the style of Taylor Swift. But the ditties still smack of soulless imitation. They’re not going to go platinum or sell out stadiums. A.I. might undermine some of the more high-volume corners of photography — stock photos, for instance — but will you use it to capture your wedding or to document a war? Nope.
Which is not to suggest there won’t still be a lot of legal wrangling. Mark Lemley, a professor at Stanford Law School who teaches and litigates on issues of intellectual property law, told me that he thinks the next five to 10 years will be marked by a series of legal battles about the role A.I plays in media. “The number of issues is going to expand and get more complicated,” he said.
Lemley is one of the lawyers representing the A.I. firm Stability AI in a copyright infringement lawsuit brought by artists. At the heart of that suit is the issue of how A.I. systems are trained. Most gain their smarts by analyzing enormous amounts of digital content, including lots of copyrighted work. Which prompts the question: Should the artists be compensated for their contributions, and if so, how?
To me, the answer is no. When a machine is trained to understand language and culture by poring over a lot of stuff online, it is acting, philosophically at least, just like a human being who draws inspiration from existing work. I don’t mind if human readers are informed or inspired by reading my work — that’s why I do it! — so why should I fret that computers will be?
Of course, one reason I might mind is if the machine uses what it learns from reading my work to produce work that could substitute for my own. But at the risk of hubris, I don’t think that’s likely in the foreseeable future. For me, a human creator’s very humanity feels like the ultimate trump card over the machines: Who cares about a computer’s opinion on anything?
Another question is how we should think about potential infringement by content produced by machines. Current copyright law is designed to protect specific creative works, not a general creative style. But if A.I. can copy an artist well enough to essentially duplicate the artist’s work, could that rise to the level of infringement? In a recent paper, Lemley and his co-authors suggested the following hypothetical: Someone sets up a site that creates “an auto-generated story about Yoda” every time a user visits. If the site charges users for the content, Lemley and colleagues argue, it could infringe Disney’s copyright — even if a human did not create the words and images about Yoda.
Such scenarios may not be hypothetical for long; A.I. generators are very good at creating near-exact copies of many well-known characters. When I asked Midjourney to show me “a small green fellow who has trained Jedi for 800 years,” its output was pretty spot on.
Which raises a related issue: Can the prompts we give A.I. be protected by copyright law? There are already prompt marketplaces, in which people sell the particular incantations they fed to A.I. to produce certain works. Can that really fly?
No doubt these questions are important, but it’s not hard to think of reasonable answers. A.I. probably shouldn’t be allowed to create direct copies of existing works, but it seems wise to allow it the same freedom to remix art that humans enjoy.
I wouldn’t allow the most common prompts to be copyrightable — you can’t claim any creativity in asking an A.I. to draw a cat — but prompts with some level of human inspiration should qualify as protectable work. After all, when I ask Midjourney to make “a cat smoking a pipe photographed on a Civil War battlefield,” I am engaged in at least a certain degree of creative work. The pipe-smoking Civil War cat was conjured by my human brain. But are such simple strings of keywords — not much more than I would enter into a search engine — enough to qualify as copyrightable creations? At the moment, we don’t really know.
But until we do, I leave you with my masterpiece:
Office Hours With Farhad Manjoo
Farhad wants to chat with readers on the phone. If you’re interested in talking to a New York Times columnist about anything that’s on your mind, please fill out this form. Farhad will select a few readers to call.