Nieman Foundation at Harvard
HOME
          
LATEST STORY
Postcards and laundromat visits: The Texas Tribune audience team experiments with IRL distribution
ABOUT                    SUBSCRIBE
Oct. 12, 2023, 2:53 p.m.
LINK: blog.ap.org  ➚   |   Posted by: Sophie Culpepper   |   October 12, 2023

Were you thinking about the applications of artificial intelligence to news in the summer of 2021? To be clear, we’re talking more than a year before ChatGPT zapped the entire internet into a new level of awareness about the tech’s potential.

I, for one, wasn’t, and I’ll wager a guess that if you, like me, were a reporter at a small local news outlet just trying to get that week’s news out (and survive), you didn’t have time or brain space to, either.

You know who was thinking about the tech and its potential for local news that summer? An organization with the bandwidth and funding to: The Associated Press. In fact, the AP had been using “natural language generation” for close to a decade, since 2014, “to automate the production of thousands of quarterly corporate earnings stories — straight from financial data feeds without human intervention.”

But as AI product manager Ernest Kung and senior product manager of AI strategy Aimee Rinehart of the AP explained in a report released in March 2022, “our suspicion has been that awareness and use of AI were not trickling down to ‘smaller’ news providers.” (The exceptions prove the rule.) So the AP launched a Local News AI initiative with funding from the John S. and James L. Knight Foundation “to understand the true state of the art at the local level and to advance smart, ethical uses of AI technology that can help local news providers augment their journalism, achieve efficiencies and sustain their businesses.” (Disclosure: The Nieman Foundation has received funding from the Knight Foundation.)

Kung and Rinehart started by surveying close to 200 newsrooms across all 50 states (and D.C., Guam, and Puerto Rico) about their AI readiness, and what applications of the tech would be most useful to them. Outlets surveyed included digital-only, print, TV, and radio-based news operations. One of the highest-level takeaways from last year’s report:

What most newsrooms and their audiences need is the automation of basic information, such as social media content and high school sports scores. They need help managing information overload. While automation does not always involve AI, it is often the first step toward the adoption of more sophisticated AI applications.

Since completing that report, the AP has launched a self-paced AI course for local newsrooms and blueprints for 15 AI-based newsroom projects. But on Tuesday, the organization announced, arguably, the initiative’s most tangible output: five AI-powered products selected from more than 40 pitches and developed in partnership with specific local newsrooms, which will all be available for other newsrooms to use for free.

Those products focus on:

  • Automating public safety incident roundups (police blotters) directly into the content management system for the Brainerd Dispatch in Minnesota;
  • Publishing National Weather Service data alerts in Spanish for El Vocero de Puerto Rico;
  • Automating video transcriptions and summaries to create a baseline structure for a news story for KSAT-TV, a San Antonio, TX TV station;
  • Building on a Minutes application that transcribes city council meetings with reporter alerts and keyword identification for Michigan Radio’s WUOM-FM;
  • Sorting community news tips and pitches into a coverage planner for TV station WFMZ-TV in Allentown, PA, which Kung called “our most complex project” in a webinar explaining the five tools on Thursday.

“It’s important to note that none of these projects are about writing reported articles,” Rinehart emphasized in the webinar. “These are newsroom efficiency tools for shrinking newsrooms that cover an ever-widening area.”

The first four projects include case studies that explain each project’s goals, usefulness to each newsroom, and challenges faced, among other details (the fifth case study will be published “in the coming weeks,” per the AP’s press release). And the first three projects include the public source code, which interested newsrooms can access on the AP’s Github.

These aren’t, by any means, the only instances of local newsrooms experimenting with AI — at Nieman Lab alone, we’ve written about other local newsrooms working on applying the tech to cover public meetings, for instance (a source of increasingly widespread interest in the local news world, and an approach now taking root in our own backyard in Arlington), and about another newsroom’s efforts to use it for comment moderation (the latter, notably, was a project developed through NYC Media Lab, which was, along with the AP, one of four beneficiaries of $3 million from the Knight Foundation to develop AI uses for news). And you may have heard about another much-mocked AI news application — Gannett’s ill-fated sports stories written by Lede AI (hot take: “a close encounter of the athletic kind” is, in fact, a timeless banger, prose to aspire to) — a crucial reminder of the tech’s limitations, and a warning sign to anyone looking to use AI for published reporting without any human intervention.

In Thursday’s webinar, though, Kung, Rinehart, and staff from the five newsrooms noted that they designed each of these products with a built-in human moderation step.

“I do want to emphasize in the era of generative AI that humans, for us, still come first,” Kung said. “It is, in fact, about AI supporting humans. It is not about AI replacing humans.”

Newsroom staff gave specific examples of how much time the tech could save them. Rafelli González from El Vocero, for instance, said automating extreme weather alerts could get them out faster than human reporters could, saving reporters half an hour of work and potentially saving lives. And Brad Rinehart of WFMZ-TV (no relation to Aimee Rinehart) said that the email-sorting tool could save his journalists as much as six hours a day, giving them that much more time to dedicate to reporting.

“We all know we have shrinking newsrooms,” said Renee Richardson, managing editor of the Brainerd Dispatch. “We have staffs that have shrunk from what we used to have. And…constantly we’re putting more pressure on reporters, asking them to do more, provide more information, add more things, do more SEO, do video, take photos, get more stories in, and rarely do we ever give them more time to do that — it’s always a crunch.”

Using AI to automate police blotters, she added, is “one project where we could actually give them time back in their day” while still getting the community the information it needs.

The AP will host a second webinar on Thursday, Oct. 26 at 12 p.m. ET, where developers from other newsrooms can meet with the technologists involved in these projects in breakout rooms to learn how to implement the tools. You can register for that webinar here.

Show tags
 
Join the 60,000 who get the freshest future-of-journalism news in our daily email.
Postcards and laundromat visits: The Texas Tribune audience team experiments with IRL distribution
As social platforms falter for news, a number of nonprofit outlets are rethinking distribution for impact and in-person engagement.
Radio Ambulante launches its own record label as a home for its podcast’s original music
“So much of podcast music is background, feels like filler sometimes, but with our composers, it never is.”
How uncritical news coverage feeds the AI hype machine
“The coverage tends to be led by industry sources and often takes claims about what the technology can and can’t do, and might be able to do in the future, at face value in ways that contribute to the hype cycle.”