Abstract
Computers are now being used to generate local news stories.
Publishers are increasingly using “robot reporters” to generate the simpler stories that used to be written by junior hacks, but can now be created by simply plugging data into a standard story template.
The argument goes that if journalists can be freed up from writing cookie-cutter stories, they can spend their time more fruitfully, holding the powerful to account.
The cynics argue that publishers, increasingly run by accountants rather than creative types, are simply trying to cut costs.
Bloomberg News editor-in-chief John Micklethwait says a quarter of the content produced by the news giant has some degree of automation.
It uses a system called Cyborg which “dissects a company’s earnings the moment they appear and produces not just instant headlines but, in a matter of seconds, what is in effect a mini-wrap with all the numbers and a lot of context”.
And it is not the only one using robot reporters. The Washington Post has developed a bot called Heliograf that automatically generates stories from sports results.
Meredith Broussard, an assistant professor at the Arthur L Carter Journalism Institute of New York University, says that while earnings reports are ripe for automation, other types of news story are not.
The potential of artificial intelligence to write more than just templated news stories is already with us. An Elon Musk-funded initiative called OpenAI has developed an AI-powered text engine known as GPT-2. Trained on millions of pages of text from the web, the engine predicts the next word in a piece of text and can write “deep fake” news stories. The stories, to the untrained eye, are virtually indistinguishable from those written by real journalists. The company has said it is so concerned about the potential for misuse of the technology that it is holding back from releasing the research in order to understand its implications more fully.
Broussard, a computer scientist turned data journalist, sits at the crossroads of AI and news reporting. Her team developed a tool called Bailiwick, which helped visualise campaign finance data related to the 2016 US election.
She says a high-profile investigative story might take years to put together and require lots of document analysis: the cost of which can stretch into the millions.
“My projects are inexpensive in comparison. That is important for innovation,” she said.
But Broussard does not believe that AI will replace journalists.
“There is this idea that technology is always the best solution – I call it technochauvinism,” she said. “There has been this idea that we can use fewer reporters and they can learn everything from social media. That has proven to be untrue.”
Many see the role of AI as a tool to help journalists rather than replace them. For those in local journalism, where newspapers are dying at an unprecedented rate, it may represent a way to avoid extinction.
Toby Abel, chief technology officer of Krzana, a company arming local news publishers with AI tools, said: “The advertising model [has been] useless for local journalism ever since social media came to the fore. We are seeing the opportunity of new funding models with the emphasis much more on the value and trust of the reader. Instead of an advertiser paying, the reader pays.
CREDIT: Gary Waters/Ikon
“The role of AI is to take the bits of journalism that are essentially busy work and try to replace those with cheap and quick solutions that get out of journalists’ way.”
Krzana – a pun in Sanskrit that means both the act of finding pearls and a human pulse – is a tool which says it is “trying to let people keep [a] finger on what is going on and finding nuggets of information”.
It draws together tens of thousands of realtime content feeds – social media, police blogs, local government minutes – and crunches them with AI tools, most of them proprietary.
“Journalists tell us they want to know whenever someone talks about a certain type of crime in Birmingham, for example,” said Abel.
He is dismissive of initiatives that aim to replace human journalists with robot reporters, saying: “The consequence has been horrible, watered-down, clickbaity journalism where they have tried.”
And it is not just large newspaper groups that are using Krzana.
The West Bridgford Wire is a hyperlocal news website, based in Nottingham, that gets up to a million page views every month.
Abel said: “It has quadrupled readership in the last 12 months through being able to source stories with just one guy doing deep investigative journalism. That seems like the future.”
Many publishers use content aggregation tools such as Tweetdeck, but The Wire’s editor, Pat Gamble, says Krzana is smarter. “It has a magical way of cleaning things up. Most things that come up are useful.”
Gamble uses Krzana on a daily basis and says a major benefit is that stories come up as they are published, rather than just when they are tweeted, allowing him to get a headstart.
And he says the site is constantly picking up new advertisers who like the fresh and relevant content it publishes.
Some of the funding for the increased use of AI in local and regional journalism is coming from two of the companies that have done much to take advertising away from local news publishers, even though they thrive on their output – Facebook and Google.
In January 2017, the Facebook Journalism Project was introduced “to establish stronger ties between Facebook and the news industry”. Its goals were to develop collaborative news products and provide training and tools for both journalists and individuals.
It has since held several hackathons (events where developers gather to create software products) attended by news organisations, some of which have seen teams develop AI tools.
Google’s News Initiative was launched the following year and aims to “empower news organisations through technological innovation”.
Abel, of Krzana, said: “I think here is a danger [the funding] gets used as a crutch to allow local journalism to continue not to make money. It should be used as a stepping stone. An injection of capital is a great way to solve a bootstrapping problem to move towards sustainable local journalism.”
One project that has received funding from Google is an AI-powered tool called Inject, which was built by a team led by Neil Maiden, professor of digital creativity at London’s Cass Business School.
Like a search engine, Inject builds an index from hundreds of sources every day and this now contains more than 10 million news stories.
“We have written a number of algorithms – you might call it creative search. Google gives you exactly what you ask for. If you put something in our tool, it takes what you write and looks for something similar but different,” said Maiden.
Inject focuses on finding an angle for a journalist to cover, whether that is data, the human angle, the long read or something quirky or funny.
To create the algorithms that drive Inject, Maiden’s team worked with experienced local journalists and politicians to rate news stories on various criteria and use the data to work out what makes a “good” story.
“We are not making the journalists more creative; we are trying to make them as creative as they would be, but faster,” he said.
Vincent Peyregne, CEO of the World Association of Newspapers and News Publishers, feels that AI offers a lifeline to some, and a new way of working for others.
“In Africa, there is strong interest in AI due to the limited resources they have. AI is an opportunity to improve and grow,” he said.
He believes the biggest challenge for wider adoption of AI is not the technology, saying: “It is about changing the workflow and culture of your organisation. The bottleneck is in management.”
Who is Using AI for News Gathering?