Machine-generated news era is not far off, but funding is insufficient.
By Jihyoung Son / Illustration by Gavin Huang.
As soon as the Korean stock market closes, a computer program collects information from the Korean bourse, puts it through an algorithm, sorts out remarkable figures — or news — and generates a 600-word straight article. All of it happens in the blink of an eye — 0.3 seconds.
The robot reporter, named IamFNBOT, is distinguished with its own byline and email address (firstname.lastname@example.org). The computer covers the local stock market and publishes an article daily for The Financial News, a local news outlet dedicated to business and economy.
From data analytics to news monitoring systems and social media-combing fintech predictors, technological advancements push journalism to new levels every year. But just how close are we to robot reporters replacing humans?
At its earliest stage, the concept of robot journalism was to boost news coverage with minimum investment. A college project in 2009 is widely perceived as the prototype of the robot writer, when a team of Northwestern University students on an academic project to tackle local newspapers’ challenge in the digital age devised Stats Monkey to generate baseball game recaps “with little human intervention.”
Stats Monkey weaves tried expressions used in sportswriting into a news story instantly, while automatically sorting match information by importance: The higher degree of unexpectedness and newsworthiness that data conveyed, the more remarkable they would be. The software later was licensed by Illinois-based Narrative Science, where two of the four students on the original team are board members.
Seven years later, robot journalism is in Asian newsrooms. More and more media corporations are trumpeting the adoption of news-generating artificial intelligence platforms. Korea’s Financial News operating IamFNBOT was one of them, following China’s Xinhua News Agency and Tencent that employed Kuaibixiaoxin and Dreamwriter, respectively.
The influence from robot journalism and algorithms to a certain business sector should be visible to attract investment. Ö The direct comparison between robot writer and human reporter comes next. -Jungsoo Kang, tech pundit and blogger
IamFNBOT, launched in January, has honed an algorithm to improve story quality. The team operating the bot, led by Seoul National University professor Joonhwan Lee, says it still requires human intervention. “We input feedback like new expressions into its software, so that it understands the context and condition for future usage by itself.” But he cannot predict when it will be able to implement a complete machine-learning process free of human support.
Lee cites prohibitive maintenance costs as the reason computers still cannot substitute human reporters. The Korean operator has faced hefty costs and a scarcity of data. “To run a robot reporter, a news organization would have to pay software developers needed for maintenance,” he says. “It would seem costly, considering wages for conventional journalists.”
At this stage of robot journalism, seed funding to back more research and development is crucial, says Jungsoo Kang, a tech pundit and blogger. The trick is to make it attractive. Narrative Science, he notes, successfully lured investors by accessing and dealing with U.S. official and intelligence data.
“The influence from robot journalism and algorithms to a certain business sector should be visible to attract investment,” Kang adds. “The direct comparison between robot writer and human reporter comes next.”
Collecting a data pool is another critical element for robots’ machine-learning capacity in terms of accurate and relevant reporting and any newsrooms willing to adopt a self-operating robot reporter.
The secret behind North Carolina-based firm Automated Insights’ news stories published on the Associated Press is big data from Zacks Investment Research, which has enabled it to generate hundreds of millions of stories a year since June 2014. Its archrival Narrative Science released its big-data-oriented tool Quill and sells its articles to media outlets, including Forbes.
IamFNBOT, in contrast, currently writes one article per day, as the source of data is “limited,” Lee claims. “I wish the team could be fed with such real-time big data. But it involves securing business contracts, which is a hard thing for a professor to do.”
The article was originally published on the website of the N3CON SEOUL 2016, n3con.com.