How (and Where) Artificial Intelligence Is Making Its Mark in Media
January 19, 2016 • Blog Post • By James Daly, WIRED Brand Lab
IN THIS ARTICLE
- Advances in artificial intelligence allow computers to analyze massive quantities of data and generate a news story
Artificial intelligence is breathing new life into the way traditional media creates and shares the news
Over the last several years, artificial intelligence (AI) has shifted from being an esoteric branch of computer science to an everyday technology that most of us carry in a pocket or purse—AI is what drives Apple's Siri, Facebook's photo-tagging, Spotify playlists and Google's auto-complete, just for starters.
But can we also expect that someday soon AI will report and write the important news of the day—and technology stories like this one? Well, guess what: It already has.
First, a bit of background: Many of the most exciting AI advances are driven by research in cognitive computing and natural language generation (NLG) processing, which allow computers to analyze massive quantities of data and generate a plain English document that highlights the most important insights. Those advances are made stronger through deep learning, a field of AI that uses neural networks to teach computers to sift through massive amounts of data to find their own patterns.
Some of the most important work with AI in media is happening at Google's DeepMind project in London, where researchers are using thousands of articles from the Daily Mail and CNN—which both feature stories with bullet-point summaries at the top—to teach computers to read and, in turn, generate stories themselves. The impact of Google's research expanded dramatically in November when Google announced it would freely share the code to its underlying deep learning engine. The wide distribution of this software (called Tensor Flow) could accelerate the evolution of AI, including getting it into the hands of software developers in newsrooms around the world.
Concurrently, commercially available natural language generation platforms—like Quill from Narrative Science and Automated Insights from Wordsmith—point to a future where computer algorithms turn mountains of data into readable stories.
The potential uses of AI algorithms in the newsroom are more than fascinating lab work—it's been used at places like the Los Angeles Times, Pro Publica and the Associated Press. “I was skeptical at first, but the more I used the program, the more I saw that what they could do," said Philana Patterson, a former business editor at the Associated Press who is now with USA Today. Patterson worked with software developers at Automated Insights to fine-tune the algorithms that translate the feel and flow of AP stories into computer code. Those programs were then applied to the wave of quarterly earnings reports.
Automating the post-game recap
AI is also sneaking into other areas of media. The popular youth sports app Game Changer creates postgame story write-ups for parents, coaches and kids. How? Algorithms (powered by AI media firm Narrative Science) synthesize a written story from all the entered game data. In a Little League game, a scorekeeper records every play. The company has 120,000 team communities as customers and is used widely for youth sports.
How do the robot-generated stories sound? Not bad is the short answer. Most stories from Game Changer might be indistinguishable from a typical write-up for a local paper. Here's the "lead" from a Little League last summer: “Will was a workhorse on the rubber for Tiburon Peninsula LL 11U All Stars. Will pitched 5 1/3 innings and allowed no earned runs, six hits and one walk while striking out four."
Reader reaction is, not surprisingly, mixed. Christopher Clerwell, media professor at Sweden's Karlstad University, recently compared human-generated sport reports against copy written by an automated program. Readers found the human-penned articles to be more enjoyable to read, while the computerized efforts felt more informative.
USA Today's Patterson believes that algorithms still can't provide the singular voice or turn of phrase that a skilled writer can, but she's optimistic that the automated system could prove useful in newsrooms that have seen dramatic staff cuts in recent years. “Staffing has gotten so tight that it can be tough to cover the some of the most basic things," she says.
Computer-generated stories could help restore coverage to areas where reporters have been reassigned, such as high school sports coverage. “Anything that can help us continue to get the product out and let reporters do more investigative, deeper and more important work should be embraced," says Patterson.
New workhorse for financial analysis
AI "news" programs are also being put to work in major industry sectors. These include places like financial services, where computers and algorithms have taken on some of the traditional work of traders, clerks and financial advisers. Alice Robbins, a manager with Narrative Science, says that 60 percent of the company's clients are in financial services, performing tasks like automating wealth management reports.
The USAA financial services company, for instance, is using Quill to read through mountains of data and add context and relevance to its financial reports. A process that can take hours with a human can be done in a fraction of that time by a machine. The report quickly tells how the portfolio is performing in a format you can read and understand.
Quill also has a role in satisfying demands for regulatory returns—the technology is used to crank out the narrative section of the suspicious activity reports for the Treasury Department's Financial Crimes Enforcement Network (FinCEN), which is used to squelch money-laundering activities.
A new platform for news ‘detection’
AI doesn't just assemble stories—it can keep newsrooms alert to breaking news. The social-news app Banjo, for instance, is an “event detection engine" that pieces together bits and bobs from various social media feeds to flag anomalies that may turn into breaking news. Instead of registering their names, users are asked to share access to their social media accounts, be it Facebook, Instagram, Twitter or China's Weibo. Banjo's social media data-mining technology now gathers information from more than 1.2 billion public social media accounts.
Companies like Banjo, Arria NLG (a UK firm specializing in cloud-based natural language generation analytics) and Narrative Science are increasingly used by organizations that need piles of data turned into readable stories. Arria works with the Met Office in London to create short weather forecast write-ups.
From a reader's perspective, AI can prove critical in helping them find content they may not have otherwise seen via paid discovery services sites like Outbrain and Taboola. Outbrain works with publishers to tease their content on related sites, as way of building traffic. Most readers have seen their tease at the bottom of article pages with a header along the lines of “If you like this you may also like…" Similarly, Taboola recommends editorial and sponsored content across many of the world's most highly trafficked sites.
A few words of caution
On the downside, AI techniques have been deployed to generate the click-bait headlines that drive traffic to some of the world's most popular websites. Lars Eidnes, a Norweigan software developer, created software that uses neural networks to write heart-pumping headlines, after training it with several million articles from sites like Upworthy, Huffington Post, BuzzFeed and Gawker.
“Neural networks are prediction machines that can learn how to map some input to some output, given enough examples of each," he wrote on his blog. “It surprised me how good these headlines turned out."
Some examples: “Barack Obama Says It's Wrong to Talk About Iraq" and “Kim Kardashian is Married With a Baby in New Mexico." Eidnes also created an entirely auto-generated news site called Click-O-Tron, which he describes as “the latest and greatest stories on the web, as hallucinated by a computer algorithm."
While few are excited about the idea of a future that consists entirely of computer-related click-bait “news," AI is destined to play a useful role in the newsroom. Ultimately, says Patterson, computer-aided algorithms are simply tools designed to help us improve how we tell great stories.
The infrastructure required to collect, process, store and analyze the future’s data requires transformational changes in the foundations of computing. Hewlett Packard Enterprise has that solution in The Machine.