Generative AI is still, for many newsrooms, in the shock and awe stage. But as editorial teams are having fun with it, many are also finding important uses for ChatGPT and its counterparts.

Die Presse in Austria is using ChatGPT for in three main ways: as a brainstorming partner, ideation and research, variation and creation, Lukas Görög, data strategist at Die Presse in Austria, told attendees at the recent INMA Media Subscriptions Summit in Stockholm.

He shared these examples:

• Interview prep: “What are 10 questions I should ask Tim Cook, CEO of Apple,” Görög said. “We’re basically using at as your brainstorming partner/assistant. You explain this is my role and this is what I want. You need to define who is asking what basically under which context. Like, ‘Please give me five suggestions for people I could interview related to the topic for my article on Bloomberg.’”
• Ideation: “I’ve copied an article and asked for five suggestions for the next article related to the topic. Maybe I get five ideas — two are bad and three are great.”
• Social media: “I need five suggestions for SEO-optimised titles/headlines and posts to LinkedIn, Twitter, Facebook, and Instagram.”

“If you want to create something, it’s more difficult,” added Tomas Bella, chief digital officer at Dennik N in Slovakia. “Results aren’t perfect, but they are more statistically right than wrong. Companies will be created this year to deal with AI laws and such. You can create fake information, but the end of the day, it’s the morals and ethics of journalists who decide what to take from it, just like Google.”

Bella shared six ways Dennik N is using generative AI in its newsroom:

• Churn prediction

• Video-to-text transcription

• Translations

• Text-to voice

• Automated article summaries

• Text-to-image

He spent most of his time talking about text-to-image, the one he admittedly has the least experience with but is the most fun.

Examples include: finding a photo of Stockholm but it’s in winter, so asking for a spring version; illustrating a photo about a brain disorder most common in older women; illustrating things that actually don’t exist, like a golden onion, a handbag made of broccoli, or a fox in a war uniform.

“We are the biggest newsroom in the country but at the same time, we are also very start-upy,” Bell said. “I went to my boss and said, ‘I need your permission to use some really bad photos for two months. And for sure, they will be horrible. AI isn’t neutral. It has biases and its own kind of style. If you just start writing keywords, you do this 50 times in a row, you see a certain style. You can override it, but you have to learn how to do that. The sooner you force your own style to the software, the better. It is willing to do what you want, but you have to tell it.”

Several problems arise with the using generative AI to create images:

Readers are confused: To Bella’s earlier point about forcing your style on the software, the more generative AI art looks like the brand, the less confused readers will be.

Graphic designers are suspicious: This is tricky, especially for early adopters. Bella suggests finding other early adopters to help ease the path.

Artists are angry: “We are trying to replace stock photos. We don’t need to save costs at the moment, so we want to keep the artists.”

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