C-3PO as a nightly news anchor? Alexa winning a Pulitzer Prize? These silly scenarios sound like the stuff of science-fiction. But the reality is that automation, which often takes the form of artificial intelligence and machine learning, is increasingly infiltrating the fourth estate and impacting how media companies gather, report, deliver, and even monetize the news.
From transcribing to fact-checking and polling to tweet parsing, artificial intelligence has been hard at work in newsrooms for years. However, the number of organizations large and small—including giants like The Washington Post, Forbes, APand Reuters—using AI and machine learning to compose content is on the rise. And that’s got the industry and consumers sitting up and taking notice.
Naturally—along with those in a number of fields—there are journalists worried about being replaced by automation. However, there are many who embrace these technological advancements, seeing them as useful assistants that help process and distribute the news.
“AI can help journalists cover and deliver the news more efficiently by freeing them from routine tasks, identifying patterns in data, and helping surface misinformation,” said Lisa Gibbs, the Associated Press’ director of news partnerships.
Chris Collins, senior executive editor of breaking news and markets at Bloomberg, agreed. “Technology is good at repetitive tasks and newsrooms tend to be overloaded with those. If you leverage technology to help with them, journalists can spend more time doing journalism—interviewing sources, breaking news, writing analysis and so on,” said Collins.
Success stories
Bloomberg built Cyborg, a program that extracts key info from corporate earnings reports and press releases. Bloomberg also has AI-assisted monitoring tools that rely on machine learning to filter out spam, recognize key names, and classify topics to cut through this noise and capture specific events relevant to Bloomberg’s financial audience.
“By doing that, we’re able to be more competitive when it comes to identifying news events,” said Collins.
AP uses a similar AI resource to automate corporate earnings articles. It also employs video transcription services that create transcripts for its broadcast customers, saving AP’s video operations personnel precious time.
Additionally, the AP’s newsroom is beginning to focus more on how AI can help the news-gathering process itself. “We recently completed a test of event detection tools, such as from SAM, which uses algorithms to scan social media platforms and alert editors when it has identified likely news events,” said Gibbs. “What we found is that using SAM, in fact, does help our journalists around the world discover breaking news before we otherwise would have known.”
Reg Chua, COO of Reuters Editorial, said his organization has been using AI for several years. “A lot of it is your basic automation stuff like scraping websites and pulling stuff off feeds and then turning them into headlines published automatically or else presenting this information to humans for checking before we publish. We also employ quasi automation and technology that scans and extracts important information from documents,” said Chua.
One of Reuters’ newest AI tools is News Tracer, which filters noise from social media to help discern fact from fiction and newsworthy angles from countless tweets and posts. “News Tracer’s core function is to tell journalists about things they didn’t know they were looking for—to quickly find news that can be reported on,” said Chua, who added that the tool provides a newsworthiness score and a confidence score to help reporters determine what to focus on.
Big and small papers benefit, too
RADAR (Reporters and Data and Robots), a London-based news service, has been a trailblazer in the realm of AI-reported local news.
“We operate as a news agency with a subscriber base of UK local news publishers,” said Gary Rogers, RADAR’s editor-in-chief. “We employ six data journalists. Our reporters work largely with UK open data, seeking out stories that will be relevant and informative for local audiences. They work as any data journalist might in finding the stories, but they use software as their writing tool in order to produce many localized versions. These are distributed to local news operations all over the UK.”
Rogers noted that AI allows RADAR to achieve a scale of story production that would not be possible by human effort alone.
“We tackle about 40 data projects each month. Each project will yield an average of 200 to 250 localized versions of the story,” said Rogers. “Since last autumn, we have been producing between 8,000 to 10,000 stories per month.”
Smaller community newspapers are investing in big machine learning capabilities, as well. Case in Point: Richland Source, a Mansfield, Ohio daily, uses a program called Lede AI to automate local sports reporting.
Lede Ai writes and publishes game recaps for every high school sporting event in Ohio immediately after it finishes,” said Larry Phillips, managing editor of Richland Source. “If it’s a big game, we will send a reporter and Lede Ai writes and publishes the first draft; our journalist adds color, flavor, and flare that can only be done by being at the game. With Lede Ai, we’ve never received a complaint about inaccurate reporting, and we’ve published over 20,000 articles.”
Education and transparency
News media professionals worry about human obsolescence in the face of such quickly accelerating automation. Yet many believe those concerns are premature or misguided.
“While this has been true in most industries and may happen in media, there is a broader picture of AI’s enabling rather than employment-destroying qualities,” Rogers said. “AI can take over repetitive and boring tasks, which frees journalists to do more important work. It can help journalists find stories by sifting large amounts of information. In our case, it allows our reporters to amplify their work, write a story in the form of a template, and produce hundreds of versions of the story for local newspapers across the UK who lack the resources to do it themselves.”
Consider, too, said Phillips, that “AI still can’t ask follow-up questions, can’t knock on the doors of multiple sources, work a beat, make a follow-up call, do the shoe-leather grunt work, garner an off-the-record comment which leads to a story angle, and certainly can’t replicate the human element, the nuance, that encompasses the very best work in the profession.”
Even if their human resources are relatively safe for now, news organizations have to navigate carefully through uncharted waters when it comes to ethics around and disclosure of AI practices.
“As these technologies evolve, having standards around transparency and best practices – such as how do we prevent bias in data from impacting our news coverage – will be critical for the entire industry,” added Gibbs.
Bloomberg’s Collins echoes that sentiment. “It’s essential to understand what technology can and can’t handle. Clearly, as with all journalism, you need judgement, best practices and processes in place to ensure what you are writing is accurate, fast and worthwhile,” said Collins. “You need to be transparent about how a story was produced, if it was assisted or published using AI. In our experience, the combination of years of human journalistic experience with technology such as AI is powerful. Obviously, the technology isn’t left to run the newsroom. It is trained and overseen by journalists, who are learning new skills in the process.”
Reading the tea leaves
Looking ahead, artificial intelligence will create exciting new capabilities as well as troubling obstacles, say the pros.
“As newsrooms increasingly embrace AI, it will help with everything from spotting breaking-news events, to finding scoops in data to audience personalization,” said Collins.
But prepare for even more fake news fiascos.
“Distribution of so-called deepfakes, assisted by AI, is a troubling trend,” Collins cautioned. “How technology evolves to both spread and combat misinformation will be a major challenge for the industry.”
Yet Richland Source publisher Jay Allred and others remain optimistic. “In the near-term at the local level, I think AI will largely be used for two things. First, it will fill the gaps on informational journalism tasks that simply are not done anymore due to shrinking payrolls,” said Allred. “Second, it will surface insights from public databases—finding out, for instance, how a particular city floods and where, how many speeding tickets were issued and where throughout a state, where do the most citations for drunk and disorderly conduct occur within a city. This will spur and support investigative journalism that wouldn’t otherwise happen.”
コメント