How will journalism evolve in the age of generative AI, as use cases proliferate along with risks?

The European Broadcasting Union undertook dozens of interviews with media leaders, academics, and other experts and wrote Trusted Journalism in the Age of Generative AI, a 191-page report, rich in details, to examine this question.

I spent hours reading the report and here, I share what I think are the most important parts you and your team should know about, as well as some interesting case studies.

What GenAI will — and won’t — change and what it brings to journalism

What will change?

The answers fall into two camps. “The first — and larger group — expects the technology to affect predominantly journalistic practices and workflows,” the report said.

Niddal Salah-Eldin, vice president people and culture at German publisher Axel Springer, foresees dramatic changes in news production: “In the long run, journalistic production will become a byproduct, more technically supported and automated.” 

This group also expects no change in journalism’s status in society and its core objectives — public service, objectivity, independence, ethics, a focus on collecting and sharing facts, telling stories, and helping different voices to be heard.

As Dmitry Shishkin, CEO of Ringier Media International, said: “I am optimistic about journalism as long as it is proving its worth to society. Machines cannot go to interesting people and talk about interesting things.”

“The second group forecasts a profound change in communication habits and, consequently, the entire information ecosystem,” the report said. 

“Some expect journalism to fundamentally change from a push activity, where media organisations serve audiences with news products, to a customer-driven pull experience, where audiences decide for themselves the format they want to interact with at any given moment.” (We at INMA forecast this in our first mini-report on GenAI in the news business as well.) 

“Today we use search engines; in the future we will use answer engines,” said Anne Lagercrantz of Sveriges Television.

So, what will remain the same? 

• Journalism is about accuracy, facts, surprise, and storytelling.

• Journalism that holds power to account will be as important as ever.

• Journalism will once again be about trusted and stable relationships with individuals and audiences

• Journalism will be about editorial choices. “This doesn’t mean that journalists will choose and curate the format or length of news items. Technology is better at this, especially when on-demand services and personalisation shape news media. But journalists will decide where to invest resources, when to push for more research and insights, and which projects to start and which ones to axe.” 

• Journalism will once again be about the real world: about meeting and talking to people, investigating, and breaking distinctive stories.

And what will emerge?

• Journalism will be about sophisticated targeting of audiences with content that enriches their lives.

• Journalism will be mainly about research, not production. The human factor in production will change and partially disappear.

• Journalism will tell stories emerging from data that wasn’t accessible before.

• Journalism will become hyper-localised in a way that wasn’t previously affordable. “New tools will help to localise news and make it accessible to audiences in new ways. Weather or traffic reports for different regions will be customised and presented by avatars. Data journalism will take general phenomena, break them down for specific regions, and generate stories that would have previously needed too much human input.”

• Journalism will become inclusive in a way that wasn’t affordable before: People can consume it in their language in a way that meets their needs.

• Journalists and audiences won’t feel the AI behind the systems and platforms they are using. This will enhance journalists’ work and audiences’ experiences.

Date for the calendar: Wednesday, July 10. Please join our monthly GenAI Webinar to hear about remarkable success stories from Tamedia in Switzerland and La Nacion in Argentina.

Use cases — and the biggest challenges

GenAI offers “an abundance of tools and opportunities to support news organisations,” the report said. 

“For example, targeting and serving different audiences with specific products and content will likely contribute to them feeling acknowledged in their needs and interests. This could build trust and increase news consumption.

“Additionally, the increasing availability of data generates opportunities to serve local audiences better with hyper-local content, including specific weather reports, real estate listings, traffic reports and the like… . For example, to serve people in different communities better and around the clock, the BBC and German Rundfunk Berlin-Brandenburg have used synthetic voices for automated traffic and weather updates.”

The BBC offers the weather forecast specific to a listener’s postal code, which “proves that text data provided by meteorologists can be converted into a rich audio weather forecast.” RBB offers app users weather and traffic updates every 15 minutes. 

AI can also make the news easier to access and understand, for example, through automated translation, text-to-speech, or speech-to-text tools, or visualising written content — making it particularly important for people with different native languages or audiences with impaired hearing or vision. 

The BBC offers a Tell Me More box that provides more contextual information within articles.

“Our prototype automatically identifies topics within an article that might require an explanation and then uses GPT-3 and pre-published BBC News content to automatically generate suggested explainable copy for those topics,” the BBC said.

The team at Brazilian news brand Agência Pública uses cloned voices reading long-form investigative stories to extend their reach, the EBU report said. 

They “had experimented with the technology when preparing an investigation into a corrupt politician who was abusing his power to push people off their land. They sent an audio version of the article, created with machine reading technology, to one of the affected farmers they had interviewed to help him respond to the investigation. This proved to be a hit — it was listened to by the farmer and others in his local community and shared with others affected by the story.” 

Finland’s national public media company Yle wanted to facilitate the integration of 62,500 Ukrainian refugees into Finnish society by providing a news service in their native language. It could launch this within two weeks, using machine-based translation, which is then checked by a journalist.

What was also interesting is what several news leaders flagged as the biggest challenge in managing AI: dealing with humans.

“(AI) is a massive topic with a lot going on all at once. People get information about developments from different sources. It’s very difficult trying to keep everyone on the same level of understanding with similar amounts of knowledge,” said Manuela Kasper Claridge, editor-in-chief at Deutsche Welle, which has more than 3,000 employees.

Jane Barrett, global editor/media news strategy at Thomson Reuters, also flagged communication as critical and said the biggest challenge she faced was prioritisation of projects: “What do we take from the experimental phase into production? Our newsroom has come up with so many great ideas. But it takes a lot of work to take something from a basic prompt, test it, integrate it into the workflow. Even more if you are fine-tuning a model or building more complex systems.”

That was echoed by Dutch broadcaster NPO’s Strategy and Innovation Director Ezra Eeman, who said his greatest challenge was “encouraging people to experiment but not put it out for production.”

For Kai Gniffke, chairman of Germany’s ARD consortium of nine public broadcasters, bureaucracy was a concern: “Getting the people who want to work with it to do so (is the biggest challenge). They shouldn’t have to wait for instructions by top management.”

Worthwhile links

• GenAI and content scraping: Looks like blocking robots.txt is not enough.

• GenAI and licensing: Time magazine signs a deal with OpenAI.

• GenAI and licensing II: ChatGPT is not linking reliably to its partners’ content, despite promising to do so.

• GenAI and licensing III: YouTube is reportedly offering record labels lump sums of cash to convince artists to let it train AI on their music.

• GenAI hallucinations: We might be able to tell when LLMs make stuff up.

• GenAI for advertising: Toys “R” Us releases a controversial, entirely GenAI commercial.

• GenAI advertising analysis: Google analysed 8,000 ads to spot trends.

• GenAI ROI: Most companies are investing heavily in GenAI but only about 35% have a clearly defined vision for how they will create business value from it.

• GenAI’s “original sin”: And how to fix it with royalties.

• GenAI for user-generated content: Like restaurant reviews.

• GenAI for staff training: Target rolls out a GenAI tool that can answer on-the-job process questions, coach new team members, and support store operations management.

A non-AI diversion

Er, is that a llama (no, not that Llama)? A llama can be your golf caddy, carry your hiking gear, help senior citizens feel less lonely in retirement homes, be part of a wilderness therapy programme for teenagers, help stressed travellers calm down, or even be a wedding guest. 

About this newsletter

Today’s newsletter is written by Sonali Verma, based in Toronto, and lead for the INMA Generative AI Initiative. Sonali will share research, case studies, and thought leadership on the topic of generative AI and how it relates to all areas of news media.

This newsletter is a public face of the Generative AI Initiative by INMA, outlined here. E-mail Sonali at sonali.verma@inma.org or connect with her on INMA’s Slack channel with thoughts, suggestions, and questions.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .