Here’s How Google’s Generative AI for Newsrooms Product Will Work
Google is pushing hard on a generative AI product that would help news outlets write, distribute, and monetize their work. “It feels very big,” one person with knowledge of the project told me. “It’s certainly a high priority project,” said another person within Google said.
The initiative has been in the works since at least last summer, but public information about it has been scarce. Last month, a report about the program alarmed me, so I went poking around for more details. And after speaking with multiple sources close to the project, I’m going to share the most comprehensive public description of the program here with you today. Let me present the facts, and you can come to your decision about the project’s merits.
Here’s everything you need to know about Google’s generative AI initiative for newsrooms:
How the tool works
A journalist first selects a “seed” source like a city council, parks department, local school, etc. they plan to cover. With a link from that single source — whether that’s a press release, 500 page report, or even a tweet — Google’s generative AI software produces a first draft of the story, complete with a lede, nut graf, quotes, and the rest. (The vision is to eventually allow for multiple input sources.) The idea is that a journalist can add to that draft with reporting and fact checking and eventually publish a full story.
What type of publications are using it
Google showed the tool to The New York Times early on, but any notion that big newsrooms might use it didn’t get very far. The company is testing it with a group of small publishers, like 10 people and under. Google is paying these publishers small sums to deliver feedback and help it build the product. Those payments will eventually dry up. It’s unclear what type of publications Google will release the tool to when it goes public, but it seems like access will be fairly locked down. I don’t anticipate wide open access. “We’re being very thoughtful and deliberate about who gets to use it,” one Google source said.
What type of journalism it’s for
Clearly, investigative journalists are not going to use this tool. Or at least, not very often. Google’s intending it for simple write-ups. “The idea is to make this a daily habit,” said the person within Google. “We’re focused on service journalism, we’re focused on very particular types of journalism, at the moment.”
More than just for news articles
The tool goes beyond helping journalists write stories. It will also help news organizations develop email campaigns and social media posts to help promote their stories. There is also a revenue component, where publishers can use it to write and send emails to their distribution lists to promote products like paid subscriptions and donations.
Many AI-generated articles are live already
You may already have read an article generated by Google’s tool. “Many, many, many articles have already been published, to no great backlash,” said one person with knowledge of the product. The quality of the output is not quite there yet for regular use, but “it’s very close.” the person said.
The disclosure decision
Google does not require publishers disclose when stories are created with the help of AI, but the tool does have a feature similar to Gmail’s “email signature” that allows them to add a disclosure to the bottom their articles.
The aggregation temptation
For publishers, a tool that can produce story drafts from a single link will be tempting to use for pure aggregation of competitor stories. People within Google are wary of this use case, but the company has not yet blocked links from competing news sites. That may come before release.
Signals within the draft
Google’s tool highlights areas where it took information verbatim from original sources, and also connects sentences in the draft to the original material they’re grounded in. The signals, color coded, will give journalists a places to zero in as they write the full story. “It’s essentially pointing to journalists and saying — These are the places you need to do the types of activities you’re used to doing: Fact check that particular sentence; ensure that if it isn’t grounded in the source that you provided, it’s still credible, if it’s not edit it, remove it, revise it,” said the Google source.
The big question
Ultimately, this project will be judged on whether it helps fill the web with better stories by either freeing up journalists to report more deeply or enhancing the quality of basic posts. There is a risk that it fills the web with AI-generated junk, which then competes with human-written stories for ad dollars and attention — and degrades the search experience — and Google is aware of this risk and trying to mitigate it. Whether it can succeed is still an open question. But we’re likely to find out soon.
The post Here’s How Google’s Generative AI for Newsrooms Product Will Work appeared first on TheWrap.