Robot Journalism Still Doesn't Sound So Scary
Narrative Science's claim that its algorithm-driven journalism will one day win a Pulitzer has human journalists quivering, yet we're still not convinced it's all that threatening to the future of journalism. Upon first looking at Narrative Science, a company that programs computers to do journalists' jobs, about a month ago, the not-very engaging text didn't scare us. Yet co-founder Kristian Hammond has given journalists cause for worry, telling Wired's Steven Levy in this month's issue that his robots will win that most prestigious prize in the next five years. He also predicts that bots will write more than 90 percent of the news in 15 years. Those predictions are indeed scary. It's a scary prediction, mostly because it's a scary vision for what Hammond thinks journalism is.
Hammond's robots take complicated data sets and turn them into stories, which appear on news sites like Forbes. So we get something like this write-up about The New York Times' earnings. The story was erroneously time stamped a full two days before the earnings were announced, but it was posted very soon after the company sent out its press release. What Narrative Science accomplishes in speed, it sacrifices in interest or analysis. Here's what the robots wrote:
What to Expect:
Analysts are expecting earnings of 2 cents per share, exactly the same as a year ago.
The consensus estimate is down from three months ago when it was 6 cents, but is unchanged over the past month. For the fiscal year, analysts are projecting earnings of 65 cents per share.
Analysts look for revenue to decrease 11.7% year-over-year to $500.3 million for the quarter, after being $566.5 million a year ago. For the year, revenue is projected to roll in at $2.08 billion.
Trends to Watch For:
For the last four quarters, the company has reported revenue decline. Revenue in fourth quarter of the last fiscal year was $643 million, a drop of 2.8% year-over-year. Revenue dropped 3.1% in the third quarter of the last fiscal year, fell 2.2% in second quarter of the last fiscal year and 3.6% in the first quarter of the last fiscal year.
That's useful information for harried bloggers to grab and create something useful out of, yes. But, also boring. There's "context," in the way that Hammond defines it: The machine knows The Times is a company. But no real context, or analysis, or prose, unlike this post human Joe Pompeo wrote over at Capital New York, where he explains what the numbers really mean, putting them in the context of the company's CEO search, or our own Alexander Abad-Santos, who judged the advertising revenue decline to be the most important part of report.
Hammond claims that as more of our lives turn into data, the robots will get better at taking that data and turning it into a story. "Humans are unbelievably rich and complex, but they are machines," Hammond said. He gives the example of baseball, which has turned more data-centric with cameras tracking every move. The data tells stories people can't, explains Levy. "Maybe the manager failed to detect that a pitcher was showing signs of exhaustion several batters before an opponent’s game-winning hit. ... This is stuff that even an experienced beat writer might miss," he writes. "But not an algorithm." But, with all the data in the world, robots, no matter what Hammond thinks, can't -- at least not from what we've seen -- put it into the most human context.
Ultimately, whether or not journalists (and the people who rely on them) should fear a machine-powered future of journalism depends on what you think journalism is. There are whole businesses built on the idea of producing massive quantities of news stories, quality controlled by machine-like formulas. Narrative Science may one day put a lot of these journalists out of work. But when most people talk about journalism, they're not thinking about rote earnings reports or baseball game recaps. (Certainly no one goes into journalism out of a passion for such things.) And shrinking one part of an industry is never good for the workforce in the rest of it.
But there is a best-case scenario -- for everyone involved -- out there. Hammond says that he thinks human-journalists will increasingly use his machine-journalism as a tool. "Maybe at some point, humans and algorithms will collaborate, with each partner playing to its strength. Computers, with their flawless memories and ability to access data, might act as legmen to human writers," writes Levy. In other words, if journalists focused less on trying to do the rote stuff that machines are better at, they might focus on producing more interesting journalism. If the threat of machine journalism ultimately makes human journalists step up their game, we'd welcome those robot overlords.
Image via Shutterstock by Charles Taylor.