A Startup Media Site Says AI Can Take Bias Out of News
Knowhere is using AI to aggregate news and rewrite it 'impartially.'
The artificial intelligence boom has expanded into creative fields once deemed uniquely human, like music, poetry, and even narrative podcasts. AI has also started writing rudimentary news articles and assisting reporters, but a new startup launched Wednesday says it will use AI to publish breaking news about a wide variety of topics.
The site is called “Knowhere,” and its creators say that they believe AI can be used to write unbiased news. The site will publish three versions of every article, aggregated from right-, left-, and center-leaning websites.
“Fake news, the Russia misinformation scandal, and all of these issues that are at the top of the Zeitgeist at the moment are all symptoms of a fundamental problem that information moves too fast and at too large a scale for us to be able to reliably parse it and understand the world as human beings,” Knowhere editor-in-chief and cofounder Nathaniel Barling and told me on the phone.
The site works by searching the internet for popular news stories. The algorithm sorts through newly published articles in near real time to determine what stories are being covered most by news sites.
Knowhere then aggregates stories from a continually expanding inventory of more than a thousand different sources with varying political persuasions to create a “knowledge graph” or database of each news story. Of course, all artificial intelligences for the moment have to have some human input: The co-founders weighed each source for trustworthiness, so a publication with a longstanding history of accuracy like the New York Times is weighted differently than a less reputable site like Breitbart.
From there, three versions of any article are published: left, impartial, and right. These distinctions are meant to show the reader how word selection and emphasis can produce biased reporting.
Barling says he believes his vision will be the future of journalism. He’s got personal ties to the industry that run blood deep. His father, Kurt Barling worked 25 year as an investigative reporter for the BBC.
“My father's main driver has always been to push me to think about editorial responsibilities and not just the technical issues,” Barling said.
Eventually, the company wants to remove the ideological labels from the stories and only offer the impartial option.
"I think the human should be in the center of it, not the machine"
While led by AI, Knowhere still relies heavily on humans besides its initial sentiment analysis programming. According to Barling, at least two human editors will review every story before publishing to check for errors, and style. Editors also check for signs of bias or favorability in the impartial versions of stories. Its current editorial team—made up of engineers and freelance journalists—span multiple countries.
“What is the absolute truth and what is absolute impartiality is almost an impossible problem if not an impossible problem so that is one of the reasons why we say nothing get published on the site unseen,” Barling said. “Everything gets reviewed, everything gets edited and everything is then reviewed post-editing as well so we have a check and balance between our journalist, our editorial team, and the algorithms themselves.”
In the end, Barling says he will give the final approval for every story.
“The buck stops with me,” he said.
Barling said the machine can aggregate previous reporting in less than a minute and can release an edited story in as little as fifteen minutes. In its current iteration, Knowhere is limited to data pulled from verified sites, which means it cannot break new stories nor report direct quotes from sources on platforms like social media. To prevent themselves from spreading misinformation, Barling said Knowhere won’t publish a story until at least five verified sources have previously written about the topic.
Knowhere is by no means the first attempt at AI integration into newsrooms.
The Associated Press has used the technology to speedily churn out earnings reports for the past three years. In 2016, The Washington Post used AI to report Olympic medal results and score updates using its in-house AI named “Heliograph,” and Reuters has recently said it wants to use AI partnered with humans to create a cybernetic newsroom.
Last year a reporter at the Atlantic even created a machine learning bot to detect whether or not Donald Trump wrote his own Tweets.
Most news sites that have boarded the AI train have done so under the condition that AI would ease the burden of tedious grunt work, leaving more room for writers to pursue more interesting tasks. With Knowhere, the machine is not just eliminating busy work but rather working alongside journalists to cover the most pressing stories.
Knowhere differs from its predecessors in both content selection and structure. Instead of reporting on heavily statistics-based and relatively low-risk stock price updates and sports news, Knowhere will write about the top trending and often most politically divisive subjects. Rather than follow a rigid template, Knowhere produces content organically through the use of neural natural language processing—a machine-learning technique which aggregates data and transforms it into a witten narrative easily digestible to human readers.
While certainly no Ta-Nehisi Coates or Hunter S. Thompson, the articles featured on Knowhere feature a writing style slightly less robotic that those that came before. In reading the site, it seems that, so far, the articles often do not differ all that significantly between left, right, and “impartial” versions.
Some, like Meredith Broussard, an NYU journalism professor and author of the upcoming book Artificial Intelligence: How Computers Misunderstand the World, remain skeptical of whether or not AI technologies can meet Knowhere’s goals.
“I definitely believe in ‘human in the loop’ systems as the future of journalism but I think the human should be in the center of it, not the machine,” Broussard told Motherboard on the phone. “It is not clear to me how this is better than anything else. They are not doing original journalism. They are just rewriting other people's original journalism and they are re-writing it with a machine, which is an interesting trick but it is not reporting.”
Knowhere received $1.8 million in funding from venture capitalists. Looking forward, Barling said they are open to multiple forms of monetization strategies, including subscriptions, paid advertising, and selling of article metadata.
When asked whether these technologies could be a death blow to human journalists, Barling responded reassuringly.
“This is about building the real future of journalism of which we are all a part,” he said. “We need to build tools to better understand the world around us given the radical changes are taking place.”
Correction: This story previously said that Knowhere was open to monetizing in the future by selling reader metadata. Knowhere said it's open to selling article metadata, not reader metadata. The story also misidentified neural natural language processing as natural language generation. Motherboard regrets the error.