The National Film Registry has collected a list of the most significant American movies in history, selected by a panel of experts with decades of scholarly experience. In a recent study at Northwestern University, a computer program crunched public online data to generate almost exactly the same list—no human expertise required. By analyzing metrics like box office earnings, Wikipedia citations, and Google’s search rankings, this algorithm was able to best human movie critics at their own game.
The study, published in the Proceedings of the National Academy of Sciences, suggests that we can use quantitative methods to arrive at what we usually see as a qualitative judgment of artistic accomplishment. Similarly, a 2013 study at Stonybrook University found that with this technique of “statistical stylometry”—a kind of sabermetrics for culture rather than baseball—an algorithm can predict with 84 percent accuracy which novels will become commercially successful. Companies like Epagogix and proponents like former statistics professor Vinny Bruzzese offer predictive analysis for screenplays as well, attempting to use equations to divine just which scripts will perform best.
It’s easy to imagine that these algorithms are killing creativity, leading us on a path to a device that spits back perfect books and film scripts like a literary version of Google’s self-driving car. One can almost hear the gathered screenwriters and novelists of the world crying out in anguish as they contemplate their future irrelevance. Are artists going to be replaced by soulless computer programs?
Software might be taking on a larger role than ever in cultural creativity, but fearing its impact is not unlike fearing that Gutenberg’s press would kill the art of writing in the 15th century.
The paranoia is understandable. Algorithms already govern most of our online experiences, from selecting the stories that show up in our Facebook news feeds to deciding prospective dates on OKCupid. They’re “taking over the world,” as one TEDx talk put it. Yet the larger fear of mechanization, the terrifying notion that creative art can be technologically replicated at the push of a button, is unfounded. I would argue that rather than replacing us, computational algorithms are actually giving us new tools to reflect on our own very human culture.
Even as a journalist writing articles that algorithms might soon replicate automatically—the Associated Press is already using software to generate its reports on corporate earnings—I’m not particularly afraid of my own creativity becoming useless. That’s because algorithms are only able to deduce trends from the data they are given, deciding which new books might be marketable or which pieces of news compelling by judging them against previously successful examples. They lack human insight.
Given this inability to “make it new,” as Ezra Pound put it, algorithms are stuck in the past. Large studios and publishing houses relying on predictive analysis could find their work suddenly derivative as they attempt to make only the safest computer-aided bets possible.
Using data for inspiration is not a new practice. Amazon allows users to vote on which TV pilots it produces. Netflix likewise uses information on what subscribers already watch to plan its own original shows—House of Cards was made because viewers loved The Social Network, which David Fincher also directed, as well as star Kevin Spacey. These strategies mimic the effect of predictive algorithms without necessarily using them. Writers have not died out nor has the world ended, though television may be a bit more boring as a result.
Likewise, Marvel and Warner Brothers are planning dozens of new superhero movies between now and 2020. The last batch of big-budget galactic brawlers was successful, so the next dozen movies will be, too, or so the thinking goes. There might be a glitch, however—to be effective, algorithms must also change with their audiences, who might eventually suffer from unforeseen hero exhaustion.
While catering to marketability might go against the sensibilities of some writers, it’s not a sin. Successful predictive algorithms could also mean that studios make stronger bets on their indie films even as they reduce risk on their mega-budget titles. There could be more room to please mainstream and niche audiences alike, plus foster the rare unprecedented hit like Frozen, Girls, or Jeff VanderMeer’s The Southern Reach Trilogy.
Algorithms are no magic bullet. Sure, Hollywood might know that audiences like quirky sitcoms with two female leads, but who are those characters going to be and what are they going to say? The computer can’t write a full script, nor can data provide all the answers. We should embrace the ways algorithms can help us rather than focus on how they might hurt.
Algorithms are simply a jumping-off point where the real creative work begins. Technology-savvy visual artists and designers are already showing how algorithms can augment creativity rather than replace it, using equations to break new creative ground. In fact, computers might be uniquely suited to finding just what makes art compelling. “Algorithms are designed for processing human culture,” the London-based artist Matthew Plummer-Fernandez says.
Plummer-Fernandez’s work includes abstract sculptures and geometric vases created by repeating algorithms until shapes are formed that no human could ever imagine. The pieces show us what it looks like when artists and machines collaborate. “Algorithms are deeply hybridized with human activity, from their programmer’s intentions to the human-generated data they feed from,” Plummer-Fernandez says.
Last November, Internet artist Darius Kazemi created National Novel Generation Month, an algorithmic response to the annual novel-writing challenge of a similar name. Instead of writing a novel, Kazemi urged programmers to design algorithms that popped out not just an idea for a television show but a completed manuscript. Entries to NaNoGenMo included a mysterious faux-Medieval tome by Liza Daly and a noirish detective comic book from Greg Borenstein. Each time one of these artistic algorithms runs, a brand new book emerges, utterly unique and impossible without the help of an equation.
The new books don’t make perfect narrative sense, hence the enduring need for a human editor. But such is the fatal flaw of algorithmic culture. It is actually fundamentally uncreative. Every set of crunched numbers, every calculated outcome, needs an equal and perhaps opposite human component to bring it to life.
The future is not so dark as algorithm paranoia might suggest. So cheer up, assembled novelists: Software might be taking on a larger role than ever in cultural creativity, but fearing its impact is not unlike fearing that Gutenberg’s press would kill the art of writing in the 15th century. New technological tools have a way of engendering more human creativity, not less.
Disruptions is Kyle Chayka’s weekly column for Pacific Standard about personal technology and the way it influences our daily lives.
