Seth's Blog : Kinds of stealing

Kinds of stealing

Most cultures are built around the idea of private property. Grain is harvested once a year and stored for months. If someone steals the grain (which is difficult to securely lock up), it threatens the livelihood of the farmer and the stability of the community. By the time Hammurabi codified Babylonian law around 1750 BCE, grain theft was already so universally recognized as a crime that the code specifies not just the crime itself, but elaborate provisions about liability in special cases.

On the other hand, by this time, folk stories about Nasruddin, the Sufi trickster, were widespread. A baker hauls a neighbor before the judge, saying that the man had stood outside the bakery each day, smelling the delicious odors of his breads and cakes without paying. The judge demanded that Nasruddin hand over some gold pieces. Shocked, the man did as he was told. The judge jangled the coins, let the baker hear the gold in exchange for the sniffed odors, then handed the pieces back.

Modern life is filled with bakery moments. Shepard Fairey artworks on the sides of buildings, music coming from the club down the street and hearing about the best play at football’s Superb Owl without paying to see the game.

Corporations don’t like this. The Cubs sued the Wrigley Rooftops to prevent people outside Wrigley Field from seeing a baseball game. The NFL doesn’t even like it if you type the name of their championship game. And since the invention of radio, ASCAP and BMI have worked hard to make sure pubs and stores don’t play songs in their venues without paying a license. Corporations get most of the benefits of copyright, not creators.

There’s another kind of theft worth understanding–this is the competition that happens when someone creates an alternative that prevents a future sale from happening. The inventors of desktop publishing stole decades of revenue from hand typesetters. The steam shovel stole job security for ditch diggers. And AI is clearly going to be making many tasks humans used to do obsolete.

When the web showed up, it first destroyed the DVD-ROM business I was building, then took a big chunk out of the almanacs, pop culture books and other detailed references that were my specialty.

[And of course, there’s the theft of trust and attention that happens all around us, every day. Attention is probably worth more than grain for many of us, and trust is priceless. Yet our culture seems to view this as a cost of being modern, not a fundamental problem.]

But the most confusing cases aren’t about spam or hustles, or even technology replacing labor – they’re about creators building on what came before. When a comedian steals a joke, is that okay? What if it’s not the whole joke, just the rhythm of it? What happens if the comedian works in a language that the original jokester doesn’t speak? Dani amplifies the style of Tamariz and Green. Is that okay?

Is David Mamet’s style of dialogue off limits to every playwright who will follow? What about Jill Greenberg’s style of photography lighting?

Is there a difference between someone intentionally using a bass guitar progression or a movie director’s style or a color palette vs. stumbling onto one without intent? Why?

Years ago, I was working with Harry Harrison (author of the story that led to the movie Soylent Green) on a science fiction computer game project. At the time, I had just published one I had created with Michael Crichton. Harrison told me that he wasn’t speaking to Crichton. “Why?” I asked, sure that a good story was about to follow. It turns out that Harry had worked for a year to write a techno-thriller about a virus that comes from space and starts killing people. A few days before he was to turn it in, The Andromeda Strain came out and was a big bestseller. Crichton had pre-stolen his idea!

As more and more of us build, spread, buy and sell ideas, we’re going to have find a shared understanding of theft. Lumping them all together makes the term almost meaningless, and diminishes the original protections on finite physical goods, which still matter.

The debate about AI training on copyrighted work is getting tangled up in the wrong question. Critics claim that because AI models were trained by reading millions of books, articles, and images, they’re committing theft at scale. But this confuses the input with the output, the learning with the creation. [N.B. The AI leaders have made a lot of dumb mistakes and selfish choices. Pirating books they should have purchased, wasting money and power, being careless or cavalier about mental health–these are all critical issues, and I’m not minimizing them. AI is changing our world, and they’re being cavalier and careless.]

Copyright has never existed to prohibit the act of reading, viewing, or learning from existing work. Every artist learns by studying what came before. Mamet watched countless plays. Greenberg studied other photographers’ lighting. The question isn’t whether AI learned from copyrighted material – it’s whether what AI helps create diminishes the incentive for humans to create.

If AI-generated content competes with and replaces human creativity to the point where people stop creating because they can’t make a living, then we have a problem. Not because of how the AI learned, but because we’ve broken the system that uses profit to encourage human creation.

The purpose of copyright isn’t to control every exposure to creative work – it’s to ensure we have enough creators. If someone uses AI trained on a million novels to produce something genuinely new, that’s not grain theft.

As usual, our culture and our technology makes everything more complex. Especially when the speed of change continues to accelerate.

Grain theft is immoral. Improving the culture is imperative. There’s always going to be tension between the two.