US Newspaper Company Starts Using AI for Publishing

21 June 2023

The largest newspaper company in the U.S., Gannett, says it will start using artificial intelligence (AI) technology to write short versions of stories written by humans.

The company recently announced a plan to use "generative AI" to add short pieces of information to the top of each story. Generative AI tools use machine learning methods to train computer systems on huge amounts of data to produce human-quality results.

Gannett says it will test the system later this year. Human writers and editors will look at what the technology produces and decide whether it meets the company's standards for publishing.

FILE - The logo of Gannett Co is seen outside their corporate headquarters in McLean, Virginia, July 23, 2013. ( REUTERS/Larry Downing//File Photo)
FILE - The logo of Gannett Co is seen outside their corporate headquarters in McLean, Virginia, July 23, 2013. ( REUTERS/Larry Downing//File Photo)

Renn Turiano is a top official with Gannett. He explained that AI technology can be useful to reduce "tedious" jobs currently performed by journalists. He noted that Gannett will test its AI system slowly before it is permanently deployed.

Some news organizations launched the technology too quickly and experienced problems. This included cases where AI systems created factual errors in stories. Turiano said at Gannett, "We're not making that mistake."

Other news organizations have said they are taking their time considering how to use generative AI tools. This includes the Reuters news agency, which provided the information for this story. Paul Bascobert is the president of Reuters. He said the news organization "embraces" AI. But he added that the news agency is seeking to use "responsible" methods to make sure the system upholds standards for accurate reporting.

Earlier this year, two publishers – Men's Journal and CNET – had problems with AI generated material. In January, CNET reported that a number of stories written by AI had appeared on its website without being checked by humans. Many of the stories contained mistakes. In February, Men's Journal published some AI-generated stories on its website that also contained mistakes.

Some experts say one of the problems with AI technologies is that they can "hallucinate," or make things up that are not true.

Nicholas Diakopoulos is a professor of communication studies at Northwestern University in Chicago. He told Reuters he thinks news organizations need to take their time with generative AI. Diakopoulos advises that news organizations not use such tools to publish stories without any human involvement in the process.

Gannett recently announced it was cutting 600 jobs. Some of the journalists still working at the company worry that AI will one day replace them. In early June, hundreds of Gannett workers left their jobs for a short time to protest job cuts and low pay. The labor group that represents some of the journalists said the company's use of generative AI is one of its main concerns.

Gannett, however, says the technology will not replace humans. Instead, it is designed to help employees be more effective and give them more time to do other, more valuable work. Journalists from Gannett's USA Today publication are currently examining test stories written by its generative AI system.

Gannett is also working to develop natural language generation technology. That system would write stories based on already-confirmed facts.

Reuters reports several large news organizations – including the New York Times, Washington Post and BBC – are all working on AI tools to help create stories. Those companies say they are currently testing their systems.

Bloomberg, a financial news organization, said it hopes to use AI to create business reports based on financial data.

Miranda Marcus is the head of BBC News Labs. The group is testing AI tools that aim to write stories based on previously published information. Those stories are called "explainers."

But the BBC says it is being careful. Marcus discussed the stories with Reuters and said they would never be seen by readers unless a human journalist decided to make them public.

Marcus said there are a number of ways AI could be used to help journalists. She called the current situation a "whole other universe of what kinds of stories we can tell with these tools." But she warned, "We're not there yet."

I'm Dan Friedell.

Dan Friedell adapted this story for Learning English based on a story by Reuters.


Words in This Story

generative –adj. the ability to produce something based on already-existing material

standards –n. a measure of quality that is considered acceptable or desirable

tedious –adj. something that takes a long time that is often not considered interesting or exciting

embrace –v. to take hold of something, to put something into use

accurate –adj. correct or precise; factual

hallucinate –v. to see something that is not real