The rise of artificial intelligence and automation in journalism has been front and center in the news lately, from Narrative Science co-founder Kris Hammond’s prediction that “a machine will win a Pulitzer one day” to Facebook’s decision to automate its Trending Topics feed.
Algorithms seem certain to play a growing role in the production and curation of news, but it remains unclear what exactly this trend will mean for journalism — or for the human journalists who currently produce it.
Celebrants argue that algorithms will simply take over journalism’s most menial tasks, freeing up human journalists to tackle more advanced work. Bloomberg editor-in-chief John Micklethwait, for example, called automation “crucial to the future of journalism,” and New York magazine writer Kevin Roose described the introduction of automated reporting as “the best thing to happen to journalists in a long time.”
However, skeptics fear that robots may end up replacing journalists instead of helping them. In a column last month, veteran newsman Robert Rector predicted a future in which newspaper executives simply “build robots to replace reporters” and lick their chops as “they envision a business with little or no overhead.” Hammond of Narrative Science did little to quell such fears when he estimated that 90 percent of news content could be written by computers by the mid-2020s.
The future of automated journalism is still a long way from being settled, but here are four early developments that might hold clues for what’s ahead.
Reuters and Graphiq Team Up to Automate Interactive Graphics
Last week, Reuters and the semantic technology company Graphiq announced a partnership to provide publishers access to an enormous database of interactive data visualizations, which Graphiq’s algorithms are constantly building and updating in real time.
“In news, speed matters,” Graphiq vice president Alex Rosenberg said in a recent phone interview. “Our solution is to presuppose what’s going to be covered, ingest the data, build all the visualizations, and then pair them with coverage, so that if you’re writing about crocodilian attacks because someone’s been attacked, we’ll have that visualization ready.”
Graphiq has already been providing this service to hundreds of publishers, but the partnership with Reuters will push that numbers into the thousands, said Rosenberg, who added that he expects human graphics designers will welcome the change.
“What we’ve seen is that graphics teams are swamped,” he said. “They want to do these big immersive infographics that really deserve a human designer, but they’re getting swamped by all the other requests.”
Photo by Maria Elena on Flickr and used here with Creative Commons license.
Facebook Hands ‘Trending Topics’ Over to the Bots
The reception for AI was much less cheery when Facebook fired its Trending Topics team last month and put algorithms in control of curating the feed’s top stories. The move came following allegations of an anti-conservative bias in the human curation, but it didn’t take long for Facebook’s algorithms to land in hot water of their own.
As Will Oremus explained for Slate: “The social network’s latest move to automate its “Trending” news section backfired when it promoted a false story by a dubious right-wing propaganda site. The story, which claimed that Fox News had fired anchor Megyn Kelly for being a “traitor,” racked up thousands of Facebook shares and was likely viewed by millions before Facebook removed it for inaccuracy.”
Algorithms have long played a role in content curation and aggregation online, and a recent Reuters report suggests that most audiences are quite pleased with the results. But on the heels of Facebook’s latest misstep, hopefully publishers will realize that even their most advanced machines still need some human supervision.
Check out last week’s MediaShift Podcast for Mark Glaser’s take on the Trending Topics saga.
Tronc Targets ‘2000 Videos A Day’ with AI
In June, Tronc chairman Michael Ferro told CNBC that artificial intelligence would help his company produce up to 2,000 videos per day. “Right now, we’re doing a couple hundred videos a day,” he said. “We think we should be doing 2,000 videos a day.”
Tronc’s business motivation is pretty clear: More video content means more opportunities for video ads, which tend to demand higher rates than other online ads. But as Timothy B. Lee explains for Vox, this strategy could prove to be short-sighted.
“You can’t fool people for very long,” Lee wrote. “If Tronc videos aren’t good, Tronc’s audience will figure it out and stop clicking on them. Facebook will figure it out and stop putting them in people’s newsfeeds. Advertisers will figure it out and take their money elsewhere.”
The question, then, is whether Tronc — or third-party contractors like Wibbitz and Wochit, which were featured in the New York Times last month — can churn out automated video content that’s as compelling for audiences as human-generated work. Tronc is betting that’s the case, but early returns suggest it’s unlikely to happen anytime soon. As USA Today’s Chris Pirrone told the Times: “The data came back very quickly that text-to-video alone, if you don’t touch it, consumers can quickly recognize it is not a high-quality product.”
AP and Automated Insights Tackle Business Journalism and Sports
The Associated Press became an early adopter of AI technology last year when it partnered with Automated Insights to produce quarterly earnings reports. The partnership helped AP improve its output from about 300 reports per quarter to 3000 per quarter — and the automated stories reportedly contained “far fewer errors” than those written by humans.
AP’s automated coverage has so far been limited to formulaic types of reporting, like game reports for college sports and minor league baseball. And the conventional wisdom is that it will stay that way for the foreseeable future.
“The things that can be automated will be automated,” Seth Lewis, Shirley Pape Chair in Electronic Media at the University of Oregon, told me in April. “But the things that can be automated still make up a relatively short list.”
Correction: An earlier version of this story misidentified the author of a story published by New York magazine. The writer’s name is Kevin Roose.
Ben DeJarnette is the associate editor at MediaShift. He is also a freelance contributor for Pacific Standard, InvestigateWest, Men’s Journal, Runner’s World, Oregon Quarterly and others. He’s on Twitter @BenDJduck.