Here's The Robot Journalist Program That Will Put Us Out Of A Job In Five Years

Is this a future award-winning journalist?

In a development that smacks of Terminators and SKYNET, startup Narrative Science has created a program that can turn data into a news stories -- stories that include, most humanely, inferences and an angle.

PopSci reported the development.

Whether statistics from a football game or housing sales figures, the robot journalist culls the numbers and puts a journalistic spin on them, writing stories faster than human competitors can for $10 per 500-word article.

Narrative Software is promoting the software as a tool to expand and supplement coverage in budget-conscious newsrooms.

The program has been utilized by a reported 20 clients for awhile now. The Big Ten Network, for example, has been using the software since spring of 2010 to recap their games. The articles feature human vernacular and euphemisms:

“Wisconsin jumped out to an early lead and never looked back in a 51-17 win over UNLV on Thursday at Camp Randall Stadium.”

With Narrative Science already putting out content comparable to Internet sports writers, it isn't hard to imagine a future of excellent robot journalism covering our top news stories. One of the company's founders was quoted as predicting that "a computer program will win a Pulitzer within five years."

That might be a bit of a stretch, but more likely -- and scary for those of us in the journalism business -- is the replacement of human writers with cheaper, faster machines, as seen in countless other industries over the years. Once a computer program can mimic the human element, the actual "human" can quickly be replaced.

The Water Cooler

Receive email updates on new comments!

cvsz
Email

2 Comments

Flag as Offensive

dlmcdonough on Sep 14, 2:58 PM said:

I don't think the pulitzer prediction is a stretch...remember that a few guys at the NSA developed a data-driven program that accurately flagged some of the 9/11 hijackers, before 9/11. That was more than 10 years ago. The availability of data, and the power to process it are much more robust now....if this gets traction, and these guys, and others, start feeding huge amounts of financial data, political data, economic indicators, political quotes, and other news data sources into a machine that is trained to look for patterns and connections that are newsworthy, then how long before the machine spots something significant that a human would not otherwise notice?

txpatriot on Sep 14, 4:17 PM said:

If you read the actual computer-generated news story, it's not bad, but it is much more stilted than a human would write.

One sentence taken out of context may sound good, but the article taken as a whole isn't nearly as impressive. And unlike the blog drones at SAI and PopSci, I actually followed the links back to the original article; I invite you to read it and decide for yourselves:

http://btn.com/2011/09/01/first-quarter-wisconsin-20-unlv-0/