This is an edition of The Atlantic Daily, a newsletter that guides you through the biggest stories of the day, helps you discover new ideas, and recommends the best in culture. Sign up for it here.
Last week, The Atlantic published an investigation revealing that tens of thousands of pirated books are being used to train major generative-AI programs. The list of authors whose work has been scraped includes Zadie Smith, bell hooks, Michael Pollan, and Stephen King; King wrote in The Atlantic today about his reaction to having his work used in this way. I spoke with Damon Beres, The Atlantic’s Technology editor, and Gal Beckerman, our Books editor, about the ethics of using books to train AI, broader questions around piracy, and whether bots can create work with literary value.
First, here are three new stories from The Atlantic:
“It Feels New”
Lora Kelley: Is the use of pirated books to train AI bad for authors?
Damon Beres: A lot of people have this persistent and valid fear that these programs are in some way going to put writers out of work. Even if you can’t prompt ChatGPT, LLaMA, or other such generative-AI tools to write you a perfect novel, they are still creating viable writing, and they are doing so on the back of existing work that was used without consent. From that perspective, I think that’s not great for authors, especially authors who already have a lot of concerns about their compensation and their place in the publishing business right now.
Gal Beckerman: If you’re an author and you write a book, and then that book is used to train AI, you are losing some control over the thing that you created. It’s not clear yet how much you are losing. But for-profit companies are building products after ingesting existing books. Authors have a right to feel anxious about it. It’s one thing to talk about Sarah Silverman or Stephen King, but it’s another thing when we’re talking about an ocean of people who create and don’t really make much money off of what they create.
Lora: Lawyers told The Atlantic that the legality of using such tools is still under discussion. Even if it’s not illegal, is it unethical for AI tools to use scraped novels and creative work?
Damon: I would say yes. To the extent that people have an ethical issue with something being stolen, this is the same issue. There is some legal haziness around this. The ethical standard in my view is simply: Was something taken and used for a for-profit program without permission? And I think that’s fairly simple.
Gal: We’ve long thought of fair use as taking a piece of intellectual property and advancing it in some way. There’s a precedent for how people do that and stay within legal limits. But with these AI tools, we’re getting to these elemental levels of a writer’s creativity—their syntax, how they built their sentences. And then the question is: What is fair use of that? It feels new.
Damon: Right. A lot of the fair-use precedents almost feel irrelevant to me. The technology certainly feels unprecedented.
And there is a broader cultural question around piracy. The programs are training on data that fundamentally stem from bundles of torrented ebooks. We know that pirated books are a component of these data sets. So now I’m wondering: If you are an individual who would not pirate a book, do you use a generative-AI program that uses hundreds of thousands of pirated books? It’s an ethical issue on top of another ethical issue.
Lora: Is there any amount of novels these tools could train on to create work with literary value?
Gal: I want to say categorically no. That would seem to me a very bleak vision of the future: one that was anti-humanist and undervalued the unique power of human creativity and reduced it to something reproducible by algorithms and through some kind of mechanistic process. I think it’s clear already that there is a certain kind of writing that AI is going to be able to do, and I’m sure that that will get more and more sophisticated. It’ll become harder and harder for us to distinguish between what a robot is creating and what a human is.
But novelists have an imagination that comes from being in the world, interacting with other people, being embodied. This is something that we don’t talk about with AI a lot, but being physical presences—getting feedback from other people and from the natural environment and from society—all of that leads to creativity. I don’t want to believe that a computer or an artificial brain would be able to reproduce that.
I think something that often gets lost in this conversation is that part of the satisfaction of reading a novel is one’s own communion with another human being’s mind. I feel satisfied when I sense that a book is the creative output of another human mind that to some degree is built the way mine is. At least for literary fiction or nonfiction at the highest level, I think part of the enjoyment of reading and engaging with books is knowing that other people made them.
Related:
Today’s News
- India successfully landed its Chandrayaan-3 spacecraft on the moon.
- Two of Donald Trump’s former lawyers, Rudy Giuliani and Sidney Powell, surrendered at the Fulton County Jail today in the Georgia election case.
- Tropical Storm Franklin has made landfall in Haiti and the Dominican Republic, bringing heavy floods and landslides.
Evening Read
Zero Lead Is an Impossible Ask for American Parents
By Lauren Silverman
Over the past eight months, I’ve spent a mind-boggling amount of time and money trying to keep an invisible poison at bay. It started at my daughter’s 12-month checkup, when her pediatrician told me she had a concerning amount of lead in her blood. The pediatrician explained that, at high levels, lead can irreversibly damage children’s nervous system, brain, and other organs, and that, at lower levels, it’s associated with learning disabilities, behavior problems, and other developmental delays. On the drive home, I looked at my baby in her car seat and cried.
The pediatrician told me that we needed to get my daughter’s lead level down. But when I began to try to find out where it was coming from, I learned that lead can be found in any number of places: baby food, house paint, breast milk, toys, cumin powder. And it’s potent. A small amount of lead dust—equal to one sweetener packet—would make an entire football field “hazardous” by the EPA’s standards.
Read the full article.
More From The Atlantic
Culture Break
Read. “A Better Story,” a new poem by Jim Whiteside:
“I walked through the rain to the American Bar, / thinking about the album he loved, the story he told / about his father. The way the pillowcase clung / to the smell of his hair”
Watch. The Amazing Race (streaming on various platforms) is a cozy reality show where the highest-stakes drama is reading a map correctly, editor Ellen Cushing writes.
Play our daily crossword.
Katherine Hu contributed to this newsletter.
When you buy a book using a link in this newsletter, we receive a commission. Thank you for supporting The Atlantic.