AI is coming to journalism, even amid high-profile debacles

Despite some early stumbles out of the gate, technologists and media executives seem dead set on introducing large language models (LLMs) into the news business.

The New York Times reported this week that Google has been pitching media orgs like the Times, the Washington Post, and Wall Street Journal parent News Corp on an AI-based tool that would help journalists produce news stories.

Meanwhile, OpenAI inked a $5 million deal earlier this week with the American Journalism Project, which will give local news outlets more access to the startup’s GPT-4-based API. The Associated Press announced a similar partnership with OpenAI last week, in which the newswire will get access to the tech company’s tools in exchange for licensing some of its archives to OpenAI, likely to be used to train its LLM.

Google spokesperson Meghann Farnsworth said in an emailed statement that Google is in the “earliest stages of exploring ideas to potentially provide AI-enabled tools to help journalists with their work” in partnership with news publishers.

“For instance, AI-enabled tools could assist journalists with options for headlines or different writing styles,” Farnsworth said. “Our goal is to give journalists the choice of using these emerging technologies in a way that enhances their work and productivity, just like we're making assistive tools available for people in Gmail and in Google Docs.”

Embarrassing errors: Despite talk of the tool as an assistant, some early newsroom experiments with LLMs have not been encouraging. Tech site CNET’s forays into AI-produced journalism last year led to a glut of copy riddled with inaccuracies, as Futurism reported.

More recently, G/O Media, the publisher behind outlets like Gizmodo, Jezebel, and Quartz, published a handful of AI-generated stories that similarly contained basic factual errors, drawing an outcry from staffers. G/O editorial leadership told Vox this week that the company has no plans to stop, despite the backlash. The interview even drew pushback from a Google spokesperson on Twitter, who disputed the G/O Media exec’s contention that the search engine favors AI-generated content.

History of automation: As the AP points out in its OpenAI deal announcement, it and other publishers have been using forms of automation for nearly a decade to mass-produce more formulaic stories, like sports recaps and earnings reports. Such formats are mostly plug-in templates for revenue figures or scores, however.

LLMs are a different beast: They can create copy from whole cloth while providing little to no transparency as to why the words were chosen, not to mention offering zero accountability when it comes to accuracy or originality. Media and tech executives seem determined to figure out how that tech fits into a newsroom, but the process so far has been messy.