News outlets are grappling with recent “generative” AI tools like Bard from Google and ChatGPT from OpenAI that can write human-sounding text on any topic based on simple prompts and questions. Some news publishers have already employed the bots to speed up their ability to write lots of content quickly, spurring anxiety and anger from human writers. But the tools still make up false information and pass it off as factual, something AI experts say is an inherent part of how the technology works, raising doubts whether it can ever be trusted to write news stories.
“We have seen large-language models like ChatGPT and Bard produce factually incorrect information. Unleashing these models in the critical, and often time-crunched, field of journalism seems premature,” said Hany Farid, a computer science professor at the University of California at Berkeley and a member of the Artificial Intelligence Lab there.
Jenn Crider, a Google spokesperson, confirmed the company was in discussions with news outlets with a focus on small publishers. The tools could provide different options for headlines or writing styles, with the goal of speeding up and improving how journalists work, Crider said. She compared the tools to AI features the company is adding to Gmail and Google Docs that automatically write emails, resumes or memos based on short prompts and questions entered by a human.
“Our goal is to give journalists the choice of using these emerging technologies in a way that enhances their work and productivity,” Crider said. “Quite simply these tools are not intended to, and cannot, replace the essential role journalists have in reporting, creating, and fact-checking their articles.”
The New York Times earlier reported Google pitching its tool to news outlets. The news-writing tool is code-named “Genesis,” and Google has had discussions with representatives from the Times, the Wall Street Journal and The Washington Post, the Times reported. A spokesperson for the Journal did not respond to requests for comment. A spokesperson for The Post did not have immediate comment.
New York Times spokesperson Charlie Stadtlander declined to comment on whether the Times has had discussions with Google, referring instead to a memo Times deputy managing editor Sam Dolnick and chief product officer Alex Hardiman sent to employees on June 7. “We recognize the power, the potential and, importantly, the risks of generative AI tools both for the public and for journalism,” Dolnick and Hardiman wrote. “We also intend to stay at the forefront of identifying creative ways to deploy generative AI to advance our journalistic mission.”
The new crop of generative AI products have sent a shock wave of anxiety through content-producing industries like marketing, film, art, music and news publishing. The bots, which have been trained on billions of words of text scraped from the open internet, are able to create human-sounding text based on simple prompts.
The generative AI tools are also trained on content taken from the news outlets themselves, without payment or permission. A Post analysis of a data set used to train an earlier version of ChatGPT showed that news stories from the New York Times, the Los Angeles Times and The Post were major sources of training data for the bot. news outlets are part of a growing movement of content creators who argue that AI companies need to compensate those whose data they use to train their bots.
Last week, the Associated Press agreed to license its news archive to OpenAI in a deal that also gave the news organization access to OpenAI’s technology. The AP has been among a group of news outlets that have experimented with writing automated articles for years.
Google has a complicated relationship with the news industry. As the company rapidly grew through the first two decades of the 2000s, it gobbled up huge portions of the advertising industry, decimating the news business globally. Local and regional papers that had relied on classified and local ads for decades saw their revenue crater, and thousands of them have shut down in the United States alone, leaving many towns without a news source beyond social media. Larger outlets pivoted toward online subscriptions, trying to avoid relying on an increasingly small share of the advertising market.
At the same time, Google search traffic is a lifeline for many news publishers, including ones who have subscription businesses. News outlets compete every day to have their stories show up higher in Google search results. Google has also been accused for years of cannibalizing traffic to news outlets by showing portions of articles directly in search results, a practice the company says helps its users.
For years, Google has tried to improve its reputation among news outlets by giving grants directly to local news and smaller publishers, as well as creating free tools like transcription software for news outlets to use. In some countries, governments are passing laws to require Google and Facebook to pay news producers directly for showing their content or portions of it on their platforms.
In Canada, a new law set to go into effect at the end of the year that forces the two tech giants to make payments to news outlets has become a major political flash point. Google and Facebook said they would block Canadians from sharing links to news outlets, while the government of Prime Minister Justin Trudeau has accused the tech giants of “bullying tactics.”
