AP expands its computer-written news program. Should reporters be worried?
Automated Insights' software can automatically generate data-driven stories on sports, business, and, soon, other industries.
Writing basic stats from a sporting event or the latest stock prices can be a tedious assignment for journalists. Take this wire for example:
“Neothetics Inc. (NEOT) on Thursday reported a loss of $6.9 million in its first quarter. The San Diego-based company said it had a loss of 50 cents per share. Neothetics shares have decreased 16 percent since the beginning of the year.”
This is important information for someone with stock in the pharmaceutical company Neothetics, but not very useful to a mass audience. These basic stories can become a waste of time and resources for news outlets, especially a wire service such as the Associated Press. But there is something very special about this seemingly normal article: a computer wrote it.
A company by the name of Automated Insights (Ai) has been working with media companies such as AP and Yahoo! to automatically generate data driven stories using an algorithm called Wordsmith.
“We can take raw data about anything – from sports to finance to business intelligence – and we take that data we have and that produces narratives written in plain language, [and is additionally] written in the tone, style, format, and length that is appropriate for that implementation,” says James Kotecki, manager of public relations at Automated Insights. “So it can be anywhere from the tone of a straightforward AP financial reporter to the tone of a snarky, joke-cracking sports reporter to an encouraging gym buddy and anything in between.”
Automated Insights has been working with AP Sports for several years, and just agreed to expand its partnership with the wire service to cover athletic events such as Division II and III football and basketball, as well as Division I and II baseball and Division I women's basketball. Wordsmith will begin automating these topics within the next 20 months. Kotecki says the AP has used automation to expand its coverage from 300 quarterly earnings stories to about 3,800.
“And for the AP that means, for example, they are able to have much more comprehensive coverage for these stories. So AP customers, which would be local papers or radio stations or what not, that may not have been able to previously get coverage about a local company that is publicly traded can now get that coverage,” Kotecki says, adding that Yahoo Fantasy Football is another example of expanding coverage. “You wouldn’t pay humans to do a million stories a week for specific fantasy football users because each of those stories has the potential audience of like two people.”
Anything that can produce data has the potential to be turned into a Wordsmith written story. It could possibly report on the tracking chips the NFL equipped its players with late last year. As Kotecki notes, the real-time information that the chips deliver about players' movements, including velocity, collisions, and touchdowns has the potential to be utilized for real-time commentary and coverage.
But sports and finance aren’t the only things Ai has seen potential in covering. As long as the company can collect the right stats about a situation, there is potential for a machine to write or contribute to an article. Kotecki says the software could theoretically even follow a reporter into a war zone, if the company could collect enough data about the area.
But are news outlets ready to hand over such reporting tasks to computers? Duy Linh Tu, professor at Columbia University and author of ‘Feature and Narrative Storytelling for Multimedia Journalists,’ says the media doesn’t have much of a choice.
“I think journalists in general, as a rule, have a hard time adapting to new technology,” Mr. Tu says, noting the slow adoption of the Internet, social media, and live streaming as examples. It can be a scary thing having an algorithm write articles, “but also it saves a lot of time, and frankly, a lot of those stories I don’t think most good journalists aspire to do.”
Tu explains that the journalism writing formula is easy for artificial intelligence to replicate, especially for stories that only involve basic statistics. While he believes the industry may take a few years to get its act together, the “desire to keep the lights on” will force reporters “to build some efficiencies,” which includes algorithms doing what many consider the “junk work.”
But both Tu and Kotecki agree that journalists should not see Wordsmith as a replacement for human reporters, but to look at it as an “ally.”
“I think there’s always going to be a need for human journalists,” says Kotecki. “We haven’t replaced a single job in journalism or any of the other industries that we’ve worked in [to our knowledge]. Instead, it’s really been not 'man versus machine,' but 'man plus machine.' ”
[Editor's Note: This article has been updated to list the full coverage of sports Automated Insights will be covering with the Associated Press.]