Never mind Jeff Bezos: “Robots” almost took over the Washington Post!
Here’s an update on a post from two years ago, “What will you write when computers write all the game stories?” That post commented on a New York Times feature about Narrative Science, a startup that turns data sets, like box scores, into narratives, like news stories.
This week we learned from this Washington Post video that the newspaper considered using Narrative Science to cover high school sports last year.
It didn’t happen, and it’s not clear from the video or this Poynter.org report why not. Was it because the technology didn’t meet the Post’s standards, or was it just a business deal that didn’t happen? Either way, Matt McFarland, who edits high school coverage at the Post, says in the video that the paper might use the technology in the future.
“The future is already here,” science fiction author William Gibson famously said, “it’s just not evenly distributed.”
Just a few years ago it might have seemed far-fetched to think that a set of algorithms could replace human writers for simple reports based on data sets. Now, Narrative Science churns out copy for the Big Ten Network, Forbes, ProPublica and others. And, as video host Brook Silva-Braga points out, they’re not alone. He mentions that his fantasy baseball league’s reports are written by another, similar company.
Tellingly, the idea that an MLB beat writer could be replaced by some software seems far-fetched to James Wagner, who covers the Washington Nationals for the Post. “No, I just don’t think it’s possible,” he tells Silva-Braga. “I think at a very basic level it’s still considered art.” Wagner does admit, though, that, knowing how quickly technology can advanced, he has reason to worry.
Northwestern professor and Narrative Science CTO Kristian Hammond dismisses the idea of replacing MLB beat writers because “it’s a crowded field.” Why go after a market with lots of players when there are so many markets that are wide open? The video mentions Little League game stories personalized for individual players as an example.
“Why would we?” Hammond asks rhetorically about trying to replace MLB beat writers. “We might make some money, but we’d make money on the backs of other people, and that’s not the way we want to do our business.”
That’s nice, but this is the real world, and if there’s money to be made, someone will try to make it. Jeremy Gilbert, also a Northwestern professor—Narrative Science started as a student project he directed called Stats Monkey—sounds a more realistic note.
“The sad truth of it is it could be financially worthwhile even if you don’t do a better job,” he tells Silva-Braga. “If you just do a just-good-enough job, then you run into the danger of replacing great beat writers who are still doing a better job than your algorithm, but your algorithm is cheap enough and doing a good enough job.”
When that happens we can all click our tongues and agree that it’s resulting in worse journalism—which we and millions of other sports fans will spend our days clicking on.
In the meantime, let me ask you again what I asked in that post two years ago, back when today’s version of the future was still two years away:
The question for you, successful or aspiring sportswriter, isn’t what sports news organizations should do with the money they’re going to save by having game stories written by machines. It’s how you will be able to make a living writing sports in a world where the job description “person who writes game stories” is likely not to exist any longer.
What can you do that, for the foreseeable future anyway, a computer won’t be able to do?