Robot reporter races against top White House correspondent in news story battle
Thu 21 May 2015
Media group NPR is the latest broadcaster to test the journalistic capabilities of robots.
In a head-to-head the U.S. radio giant placed one of its top reporters and White House correspondent Scott Horsley against the computer software WordSmith, developed by a company called Automated Insights which programmes the artificial intelligence (AI) system to produce hundreds of millions of short news stories every year for sites such as Yahoo! and the Associated Press among others.
In the latest test both robot and human contenders awaited a financial earnings report from American fast food chain Denny’s. Once the report was received, the two were up against the clock to produce a short news story ready for broadcast on NPR radio. In addition to speed, the final draft would also be assessed on style.
WordSmith produced its story in just two minutes, while Horsley took just over seven minutes.
The end result was as follows:
Of course, the story on the left is produced by WordSmith – evidently not replicating the human style and insight of Horsley’s. The human story came up tops in an NPR poll, winning 3403 votes to WordSmith’s 270.
However, according to those at WordSmith, the tone and style of its articles can be amended with some input from the company’s engineers. The Denny’s piece had been programmed to mimic the straightforward style of AP releases, but if set to study NPR specific stories WordSmith would be able to emulate the radio group’s writing – “learn NPR’s style, and start slinging its own breakfast-food metaphors.”
The NPR human vs. robot reporter test comes more than two years after The Los Angeles Times started testing the strengths of robo-journalism. The newspaper published a story on 1st February 2013 about an earthquake, produced solely by an algorithm programmed to generate short articles as soon as an earthquake is recorded in the area.
The LA Times’ technology draws on data provided by the U.S. Geological Survey to insert the real-time information into a pre-written article template. It has also developed a similar system to write articles about local incidents of crime, with human editors providing input where necessary.
“It saves people a lot of time, and for certain types of stories, it gets the information out there in usually about as good a way as anybody else would […] The way I see it is, it doesn’t eliminate anybody’s job as much as it makes everybody’s job more interesting,” said Ken Schwencke, journalist and LA Times ‘Quakebot’ algorithm programmer.
The industry argues that the technology will remain as a supplemental tool, but who knows how long it will be before robots conquer the newsroom.