By Patrick White, professor of journalism at the University of Quebec at Montreal (UQAM).
Artificial intelligence (AI) already has a major impact on journalism. Robots can write basic sports results articles and basic earnings reports of publicly listed-companies for wire services like AP and Bloomberg. It is good for political campaign finances as well. At Bloomberg they use AI to produce news alerts or articles when a stock moves up or down significantly and also for hyper personalized news.
AI could be useful for the news consumption of young people : for example by talking to bots or for voice-recognition tools.
But AI is not magic. And Deep Fake videos produced by AI are very dangerous. The privacy issue is key.
The Reuters news agency even created its own manipulated video in order to train its journalists in how to spot fake content before it gets shared widely.
At the Agence France-Presse (AFP) news agency, a recent partnership with Facebook uses AI algorithms and users' feedback to fact-check news. And it works. AFP debunks Fake News stories on a daily basis with this new team.
Some big media like The Washington Post have created a Machine learning software as a tool to help journalists analyze huge amount of data.
Meanwhile, a text-generating "bot" nicknamed Tobi produced nearly 40,000 news stories about the local results of the November 2018 elections in Switzerland.
How machine learning can (and can’t) help journalists
While machine learning holds great promise for journalism projects that involve lifting patterns from large quantities of data, some data journalists warn that their colleagues are “getting too excited” about the technology, writes Floris Wu. “There are probably relatively few circumstances under which reporters are going to need to acquire machine learning — it’s really where you’ve got a classification task,” said Peter Aldhous, a science reporter at BuzzFeed News whose series “Hidden Spy Planes” relied on machine learning for data sifting. In some cases, he said, simple things like a keyword alert or standard statistical sampling techniques might just do as good of a job of parsing information, in an even shorter amount of time. “We need to use the right tool for the right job. [For much of what we do], we don’t need machine learning; we need good data reporting.”
AI can be good for personalized story recommendations for users on news sites. That is an avenue that will be explored more than ever in the near future. We have only seen the tip of the iceberg here.
The Heliograf AI system at The Washington Post helped this media to cover the Olympic Games in depth with hundreds of articles in 2016. These articles were basic and computer generated about specific athletes and sports. It is a start.
So is AI to replace journalists then? I say no because of the monotonous reporting. Who will want to spend his or her times read boring sports results or financial results in two paragraphs? Not that many consumers I believe. Consumers still want added-value in journalism.
Yes, robots will write stories about hyper local sports results and financial results, but only for cost savings and streamlining.
At The Associated Press (AP) news agency, AI actually helps to free journalists to do added value content; they also created a Working Automation group at AP. Robots also write summaries of AP stories.
The Bloomberg News case
At Bloomberg News, a financial wire service, automation is helping them write thousands of stories a week coming from loads of unused financial data. Top editors of Bloomberg presented their findings at the South By SouthWest (SXSW) media convention in March 2019. I was quite impressed by the technological development at Bloomberg.
AI also helps the Bloomberg news agency to be very fast in breaking news (traders need that) and it automates thousands of translations of their articles into multiple languages.
Bloomberg now describes itself like a data and tech company and not like a media company because of AI.
AI has helped scale the production of news and the distribution of content. A third of the content at Bloomberg now comes from automated assistance; market information is the bread and butter of Bloomberg News.
AI alerts Bloomberg journalists on major stories and data; it does data correlation but it has its limits. The journalist double checks the content; build stories around AI data.
AI detects patterns in data but cannot do all the work. But the risk of errors is lower and accuracy is key, top Bloomberg officials told a SXSW panel last March in Austin, Texas.
AI has been used to to fact checking on Donald Trump at many news agencies and media organizations.
On the social media side of things, SAM, a social media Artificial Intelligence (AI) engine, works with Snapchat and Twitter to turn snaps, tweets, photos and videos into actionable data by using algorithms. It is proving a very fast way for journalists to keep their finger on the pulse around them, helping to pinpoint and verify breaking news events.
I am quite confident that AI will not hurt journalism as much as we thought a few years ago. I strongly believe we will always need journalists to analyse the world, put things into context, tell stories, pick and choose top news and make sense of events. AI can only contribute to new ways of reporting and better use data in general. I remain optimistic.