What if a computer created your news? A computer that tailored its style, tone, and slant to exactly what its owners wanted? Actually, it would look like The Washington Post, who have been running a neural network to scrape events, create articles, and publish them for over a year now.
Heliograf, as the program is called, has been writing smaller articles for the Washington Post since the Rio Olympics. These articles have ranged from basic reporting on finance to election coverage.
What does this mean for WaPo? For one, they’re trying to move out of the publishing niche and start selling tookits directly to other outlets. For another, they want to avoid sending people to cover sports, economic news, and other pieces that aren’t strictly analytical.
Like a set of fridge magnets
How does Heliograf work? First, editors figure out what the outcomes are. Next, they input those outcomes into Heliograf as phrases. Heliograf checks in on what actually happens, then finishes, proofreads, and publishes the article.
Other artificial news reports include Wibbitz, leveraged by USA Today as a video production tool; Reuter’s News Tracer, a Twitter analysis engine; and Buzzfeed’s now open-sourced BuzzBot, which scrapes crowd-sourced data.
The common thread here? People still need to make a story, whether that means pre-making or sorting through collected data. According to Digiday, AI just doesn’t work out for personalized, interpretive news – or local news. For the foreseeable future, we at Bham Now aren’t going anywhere! Plus, who wants to interview with a chatbot?