Welcome to the latest Newsroom Initiative newsletter.

This week, I’m in Perugia, Italy, for the International Journalism Festival, one of the most inspiring and educational events in the calendar for our industry.

The Perugia event is strong on news start-ups, big on innovation, and huge on journalists from media organisations of all sizes from all over the world talking refreshingly openly about their challenges and achievements. And it’s a great place to hear new ideas.

In this edition, I’m calling out two sessions on generative Artificial Intelligence in newsrooms and another on membership models and how they fit — or don’t — with advertising and subscriptions. I’ll talk about more in later newsletters and put some of them on the Newsroom Initiative Slack.

Generative AI has great use cases in newsrooms right now

We’ve quickly become aware of the risks generative AI poses to journalism, whether it be questionable accuracy or the theoretical threat to jobs. But there are some great tasks it can help with right now — search engine optimisation for example.

In a presentation to the International Journalism Festival in Perugia, Nicholas Diakopoulos, director of the Computational Journalism Lab at Northwestern University, laid out the hazards but also opened my eyes at least to some of the instant opportunities in using some of the generative AI tools and services that already exist.

A couple of use cases he demonstrated that really struck me were document analysis and search engine optimisation. Nick described those newsroom tasks, along with content discovery, translation, tips processing, and text summarisation as “back office” in that they are helping us do the work behind journalism, not creating content to be published directly.

Handy tools

In the document analysis exercise, Northwestern researchers asked ChatGPT to go through an academic paper submitted to the Arxive online repository and return a coherent summary, extracting the key points and attributing the findings properly.

“Generative models are really good at analysis,” Diakopoulos said.

It worked brilliantly but also exposed a few additional lessons newsrooms need to be aware of. It’s all about how you write the prompt — the request you submit to the generative AI interface. It is clear that prompt writing may in a sense become a skill in itself, almost a mini-algorithm to explain to the uber-algorithm what it is you are looking for, what it might be used for, and suggestions on where the engine might look or compare your request with.

Diakopoulos described it as “how to deliberately control what the model should do and shape the outcome.” He talked of either a “zero-shot” approach (where you ask a single question about what you want) and a “fine-tuning” approach (where you write the prompt in a way that gives the engine a much clearer sense of what you are looking for or will do with the results).

The experiment which I tried and really hit home to me was asking the ChatGPT engine to suggest search engine optimisation terms that would suit a given piece of copy. Having written an INMA report on Google search last December, I am very aware of how time consuming it can be to do SEO well and how critical sensible terms are to being discovered. (Yes, there is an irony also in using generative AI, which some see as the future of search engines, to maximise your reach in the currently dominant search engine.)

Working with a friend, Catarina Carvalho, founder of Lisbon city site Mensagem, we wrote a prompt asking for suggested SEO terms to support a Mensagem story. The engine almost instantly returned an impressive set of search terms that even a cursory look suggested would be ideal to promote the story and which certainly would have taken the reporter or editor far longer and probably more arduously to create.

That was driven by posting the URL of the story in its English version. 

Interestingly, posting the plain text of the story delivered a rather different result, perhaps (and this is a guess) because the URL version has more context of other channels and the navigation on the site so maybe a greater sense of what Mensagem is all about. 

I had been wondering if ChatGPT might be good for SEO suggestions and now I know. It is potentially amazing.

You need editors

On the potential downsides or risks of using ChatGPT for more reader-facing tasks, Diakopoulos said his research suggested it was critical a traditional editor vet anything before publication — and by vetting he really means deep editing and fact-checking. Given some newsrooms struggle to edit reporters’ copy before publication, that may be a problem.

One of the biggest problems with the ChatGPT models is their propensity to generate fabricated content and links, so called hallucinations. Content generated from them needs to be checked before publication. In tests, as many as half of all content generated was wrong either in creating fabricated conclusions or giving the wrong attribution on sources.

“This might improve with better prompting, but you’re going to want to have humans integrated into the publishing loop to check content before publication, reading every sentence and asking ‘Is this true?’ or ‘Did the model hallucinate?’ You really need to edit these things,” he said.

My take: Prompt writing is going to be a huge skill in newsrooms whether among editors or reporters. Experimenting right now is critical to start to get it right. I have even found that using words like “please” and “thank you” may give a different answer.

Diakopoulus had an interesting formula to think about when writing prompts: “subjects+compositions+styles.” What is the subject you are asking the engine look at, what sort place do you want it to look in, and what type of result are you seeking? Go and play with it here in the Open AI Playground.

INMA has just released a deep-dive report News Media at the Dawn of Generative AI, free to INMA members.

AP CEO warns of “existential threat” to publishers from generative AI 

At the other end of the spectrum, the chief executive of the Associated Press said last week that generative AI engines built on content scraped from publishers pose an “existential threat” to media owners unless regulators impose laws around attribution and ownership, 

“There is no doubt it poses a serious challenge to our intellectual property, and that is a big concern to all content creators and originators of media,” Daisy Veerasingham told a session of the International Journalism Festival in Perugia, Italy. “We have to as an industry come together to create a legislative framework that will better protect our IP because these tools are learning and are becoming smarter as a result of the work that we do. We need to ensure that we are protected. Otherwise, we will face an existential threat if we don’t respond in that way.

“We can’t allow what happened with search to happen again,” she said in comments during a session on a new generation of media CEOs. 

The AP had been taken by surprise at the scale of its content that had been ingested by Open AI to help build the large language model beneath its ChatGPT engine, Veerasingham said. She urged publishers and legislators to move fast and decisively to protect the ownership of content.

In response to a question from INMA, Veerasingham elaborated on her concerns that the corpus of Associated Press — along with Reuters and virtually all other international publishers — had been absorbed by Open AI to create the model that powers its generative AI tools.

I wrote about the sources of the content and the problems with attribution in an earlier post, and my colleague Ariane Bernard, who leads the INMA Smart Data Initiative, tackled the subject of copyright in a blog published today.

“We were surprised, and that’s why I make my point that it is an existential threat if we don’t figure out how to protect everybody’s copyright because it’s about all of the content that these systems are using that will make them smarter, that will make them be able to compete with us,” she said. “We have to be honest about that they will be able to compete with us if we don’t work out ways of protecting our intellectual property.”

OpenAI has said its large language model is based on content ingested from the Internet and published prior to September 2021. It collected the information on the basis of research, but OpenAI’s shift to a for-profit model and collaboration with companies such as Microsoft means it is arguably monetising publisher content.

As with clashes between news publishers and Facebook and Google, Robert Thomson, chief executive officer at News Corporation, has led the charge to defend the interests of content creators in preserving the long-term value of their journalism and historical archives.

“When AI is being trained, they are often using professional content — i.e. our content — and we should be rewarded for that,” Thomson told a conference in Australia this month. “It will clearly be surfacing ­individual stories or features that we’ve done, and we need compensation for that … and most dangerously with AI and bigger intelligence, they will be synthesising aggregated content from various sources and surfacing a snippet.”

News Corp has said it is already in talks with a so far unnamed generative AI company. AP’s Veerasingham said publishers needed to work much more collaboratively than they had in the original days of internet search to combat the threat from generative AI. (Thomson can be expected to talk about generative AI in his keynote at the INMA World Congress in May.)

Veerasingham noted that the shift to generative AI search was a direct threat to a significant percentage of traffic to news sites: “If your biggest monetisation model is driven by advertising and search and now you don’t have to go back to you as a content originator for those results then that’s a fundamental other pressure on that business. We need the money to keep journalism funded in the world at large.”

She also made clear that Associated Press was an enthusiastic user of Artificial Intelligence and had been for a decade. That would only grow in the future as routine tasks lent themselves to automation that would unlock time for journalists to be more effective.

“There is no doubt that using Artificial Intelligence helps the efficiency and effectiveness of your news report… . It has improved our efficiency, it’s taken away a task that we were doing and allowed our journalists to focus on more valuable reporting,” she said.

Member models hit some subscription barriers 

Membership models have been a fashion among news start-ups, led in part by the now-defunct Dutch site De Correspondent, which ultimately collapsed in 2020. The idea is to gain support from passionate members rather than just subscribers and to create or address communities.

Three followers in the De Correspondent mold, however, suggested the membership model may face some of the same challenges as the traditional subscription model from fatigue with news, to questioning the value of news and of membership itself, and the cost of living.

“We have definitely taken a hit in the past few months,” said Leon Fryszer, chief executive of the German site Krautreporter. 

He and others on the panel about membership models at the International Journalism Festival in Perugia said membership was challenged by a range of factors, including a push towards subscriptions and forms of membership from more traditional publishers.

“You have to constantly retell the story, why are you here, what is the problem in the world that you are trying to fix,” said Lea Korsgaard, editor-in-chief of the Danish site Zetland. “That’s what journalism failed to do for so many years. We took for granted that people could figure out for themselves why journalism mattered.” Her solution for now was to invest more in journalism.

Fryszer said he believed Krautreporter would investigate other revenue streams and methods to reach an audience it could convert to loyal membership, especially as a general rather than local or specialist publication. In a sense, membership-led sites had pioneered the idea of paying for news at a time when readers were exhausted with clickbait. Now, however, everyone was chasing subscribers or members.

“It’s become easier for everybody — for us, it means our comparative advantage has changed,” he said. 

For much more on the whole reader evenue model, see the INMA Readers First Initiative.

Recommended follow

The official Twitter account @journalismfest of the International Journalism Festival wraps developments at the event and shares updates on what’s new in the industry. The hashtag for this year’s event was #ijf23.

Talk back

Tell me what you want to read and what you like or don't like in this newsletter, please. E-mail: peter.bale@inma.org. There’s also an INMA Newsroom Initiative Slack channel.

About this newsletter

Today’s newsletter is written by Peter Bale, based in New Zealand and the U.K. and lead for the INMA Newsletter Initiative. Peter will share research, case studies, and thought leadership on the topic of global newsrooms.

This newsletter is a public face of the Newsroom Initiative by INMA, outlined here. E-mail Peter at peter.bale@inma.org or newsroom@inma.org with thoughts, suggestions, and questions.

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