US media conglomerate Gannett plans to incorporate generative artificial intelligence into its story distribution system, at a time when news organizations are looking to capitalize on this technological revolution in an effort to save costs and improve productivity.
The largest U.S. newspaper publisher, with more than 200 dailies, said it would use humans in the process, explaining that technology cannot be automatically relied upon without oversight.
“Generative AI is a way to increase efficiency and remove some of the daunting tasks for journalists,” Gannett senior vice president Rene Toriano said in a recent interview with Reuters.
Still, he added, “the haste of some news organizations is misguided.”
“It has taken a responsible approach, It balances accuracy with increased trust,” he said. explain.
Newsrooms Battling Artificial Intelligence
Many U.S. newsrooms are grappling with how best to integrate artificial intelligence tools to generate new content or data in response to user questions.
Some experts say there is suspicion of misinformation, which is especially problematic for an entire industry that relies on accuracy.
“I would not recommend using these templates for any self-published journalistic purpose,” said Nicholas Diacopoulos, an associate professor at Northwestern University.
Gannett’s strategy reflects the deliberate approach some newsrooms follow, as its caution comes from mistakes made by CNET and the Men’s Journal, which use artificial intelligence technology to create content and stories, but contain catastrophic mistakes. .
In the next quarter, Gannett will launch a live pilot that uses AI to identify the most important points in an article and create summaries on top of them.
The feature is slated to roll out in the fourth quarter of USA TODAY, at which point journalists will have the final say on which AI to use. Eventually, Gannett will integrate summarization technology into its publishing system.
reporter’s fear
Gannett reporters are working to ensure they are not replaced by technology, and the union says hundreds have already left their jobs as a result of June 5 layoffs and stagnant wages.
“We’re not worried about replacing our teammate, so we don’t even think he’s an appropriate replacement,” said Ilana Keller, a reporter for the Asbury Park Press in New Jersey.
A spokesman for the company said its use of artificial intelligence would not replace journalists, but instead use it as a tool to help them be more efficient while continuing to focus on creating more valuable content.
Gannett laid off more than 600 employees last year but has $1.23 billion in debt since merging with Gatehouse in 2019, but has been living with the consequences of its decisions since.
Gannett is also developing a generative AI tool that can cut long-form stories into different lengths and shapes, such as bullet points or captions on images to create slideshows.
To summarize the stories, Gannett relied on OpenAI competitor Koher, the developer of ChatGPT, who spent two weeks training Koher’s large language model on 1,000 previously published stories, with summaries written by her reporters.
To further practice the model, automated summaries and bullet points were reviewed and edited by reporters on the USA TODAY policy team.
While most news organizations have long relied on some form of AI to recommend and personalize content, new developments in generative AI are sparking renewed interest in the industry.
Gannett also experimented with natural language generation, in which AI builds written text around real-world data, which creates a story but needs to be vetted by journalists before publication.
Other news outlets are approaching generative AI with varying degrees of commitment and caution. The New York Times and The Washington Post are in the planning stages, according to a memo seen by Reuters.
In turn, Reuters rival Bloomberg is developing its own generative artificial intelligence model, Bloomberg GPT, which has been trained on financial data.
Reuters editor-in-chief Alessandra Galloni said in a letter to staff in May that Reuters uses artificial intelligence to convert audio to text to help translate video because it does not publish AI Generated stories, videos or images.
The BBC News Lab also created a prototype based on previously published BBC content, written using the ChatGPT-3 model.
“The stories we can tell with these tools are a whole other world,” says Miranda Marcus, head of the BBC News Lab. But we’re not there yet. ”
(Reuters)