With newspapers shedding jobs and losing subscribers, it's a hard time to be a professional journalist. Narrative Science, a company affiliated with Northwestern University, just made it even harder, with new software that can write news articles in under a minute, using facts, statistics, earnings reports, and other data fed to the computer. The articles sound like they were written by humans, impressing robotics and language experts. Is this technology really so good that human writers are a big step closer to being obsolete? Here, a brief guide:
How does the software work?
Northwestern computer science professors Kris Hammond and Larry Birnbaum started Narrative Science to make money from a program they'd written, Stats Monkey, that creates sports stories through algorithms that use box score and play-by-play data to generate short articles. The program also produces headlines and selects photos of key players. Narrative Science now offers to write stories based on financial reports, real estate data, polling and elections, and local news, too.
How is this different that previous "robot journalist" efforts?
Unlike clunky, artificial-sounding fill-in-the-blank computer programs, Narrative Science draws on historical data and a set of concepts to shape a story around. In sports, for example, the program might decide if a game was a "group effort" or "individual effort," then pick appropriate vocabulary. If it's a lopsided game, the computer might call it a "rout" instead of a win; if one team pulls off a late upset, it might be a "come from behind" victory. "I thought it was magic," says venture capitalist Roger Lee, whose company helped raise $6 million for Narrative Science. "It's as if a human wrote it."
Is it really that good?
Customers and linguists seem impressed. Here's the opening sentence from a news brief on an in-progress college football game earlier this month: "Wisconsin appears to be in the driver's seat en route to a win, as it leads 51-10 after the third quarter." And the algorithm is getting better as more companies use the program. "In five years, a computer program will win a Pulitzer Prize," predicts Hammond, "and I’ll be damned if it’s not our technology."
Who's using the software?
Narrative Science says it has 20 customers, but won't divulge any names. Companies that acknowledged using the program include the Fox Sports–affiliated Big Ten Network and construction industry trade publisher Hanley Woods. Other clients include newspaper chains looking to expand coverage of local sports and business news at low cost — Hanley Woods pays Narrative Science about $10 per 500-word story, and that cost will almost certainly decline as more companies sign up.
Should cub reporters just quit now?
No, say Hammond and Birnbaum, who are also journalism professors at Medill. So far, the program is mostly being used to generate stories that otherwise wouldn't have been written. That may be, but given its low cost and lightning speed, it sure seems that "it's not just the manufacturing industry that's prone to being replaced by machines," says J. Angelo Racoma at TFTS. It's "up to us humans to up the ante in terms of quality" if we want to remain relevant. Robots "can already play soccer and beat us at Jeopardy," and a robot journalist "probably wouldn't need things like caffeine in the kitchen to keep it humming along," says Chris Gayomali at TIME. So yeah, "we're doomed."
THE WEEK'S AUDIOPHILE PODCASTS: LISTEN SMARTER
- How the South's ugly racial history is haunting ObamaCare
- What if Leo Strauss was right?
- 43 TV shows to watch in 2014
- If Democrats abandon immigration reform after Tuesday's likely loss, they will turn 2016 into a debacle
- Stop making fun of philosophy and read some philosophy
- How to be the most productive person in your office — and still get home by 5:30 p.m.
- Beware of Splenda: The backlash against artificial sugars
- The real story behind Deliver Us From Evil
- 6 things the happiest families all have in common
- Feast your eyes on this beautiful linguistic family tree
Subscribe to the Week