Adapt or die: How Gazeta Wyborcza uses AI in the newsroom

Boles?aw Breczko was curious whether AI is already threatening his profession. So in early April he asked ChatGPT to write a text entitled “Elon Musk vs. OpenAI: The race to create a new human-computer interface”.

The piece ran on Gazeta Wyborcza’s website with Breczko’s byline and without an explicit disclosure that it was generated by AI (Breczko says he tried to insert some hints, though they went largely unnoticed by readers). Even most of the journalist’s colleagues didn’t know that the text was written by ChatGPT – only Sebastian Ogórek, the chief editor of Wyborcza.biz, was informed beforehand. 

Soon after the initial article ran, Breczko revealed and reflected on the experiment in the article titled “Are journalists still needed? My last article was written by ChatGPT”.

“I didn’t want to mislead or deceive anyone, I believe that texts written by AI should be clearly marked. It was the only way I could see if anyone would recognize that the material was not human-made”, Breczko wrote.

The goal of the experiment was to find out if ChatGPT, after analysing Breczko’s texts, would be able to write a quality article in his style. To everyone’s relief, it failed to do so. (For the sake of the experiment, Breczko asked several journalists and editors unaware of the provocation to evaluate this text after its publication on the site. They highlighted multiple issues, including poor style and lack of context – and were relieved when Breczko revealed the text was written by ChatGPT).

However, Breczko believes it’s possible ChatGPT might master the task in the future. Already now, it may be a threat to news journalists who write short texts based on press reports and other open sources.

Defining AI do’s and don’ts for the newsroom

Talking about Breczko’s provocation, the editors of Gazeta Wyborcza underline that this is the only text written by ChatGPT on Wyborcza and its role was strictly experimental.

However, Wyborcza had begun exploring AI technologies for newsroom use already in January 2023, three months before Breczko’s publication, deputy editor of Wyborcza.pl Aleksandra Sobczak told The Fix.

“We formed a special group of 61 people that analyses various artificial intelligence tools daily and discusses this topic. Its members are our journalists, editors, programmers, and IT specialists”, Sobczak says. 

Gazeta Wyborcza now is in the testing stage with AI. Sobczak explained to The Fix that the research group has defined two rules regarding the work of Wyborcza journalists with ChatGPT: 

• Prohibition of using it as a tool to write text or as a source of information (it has to be checked in legit sources);
• Prohibition on sharing confidential content with it to keep the informants safe.

ChatGPT is allowed to be used as a tool for brainstorming, for example on unapparent interview questions or additional informational sources.

The internal AI group also considered the use of AI-generated images like MidJourney, inspired by the Slovak outlet Dennik N, Sobczak told The Fix.

By January 2023 Dennik N, a Slovak national newspaper, was using generative AI to illustrate 10%-12% of articles every day; soon after, the first AI-generated image was printed on the front page. Dennik N uses AI for generating images not because it’s cheaper than original photos or art but because it’s quicker and improves overall visual quality, eliminating bad stock photos, Tomas Bella, the organisation’s chief digital officer, said in an interview with International News Media Association.

However, Wyborcza itself has so far decided against using AI-generated images to illustrate articles on the website. Although aesthetically very interesting, services like Midjourney raise ethical doubts as they draw on the works of other artists, Aleksandra Sobczak told The Fix. However, she doesn’t rule out the use of generative AI to illustrate articles in the future.

The future of AI and journalism

Even if human journalist won’t disappear, AI brings other challenges to the media.

“Already now, using the new Bing is more convenient than Google. Search right away provides an answer to the question asked in an accessible and specific way. It’s a nice change compared to clicking through ads and empty SEO texts displayed by Google search”, Breczko told The Fix. (Google has started catching up in recent weeks, rapidly integrating AI capabilities in search, though much of its latest advances aren’t available in Europe at this point). In his opinion, besides convenience, this brings a new threat that users will blindly believe what the AI ??tells them, even though its answers may be based on incorrect information, or just be made up.

Sobczak underlines fake news issues made worse by AI. Many people, in order to find the truth, may turn to traditional journalists who care about the reliability and quality of information, she told The Fix. At the same time, trust in the media and the journalistic profession has declined for many other reasons, she adds.

Dominik Uhlig, Wyborcza’s data department BIQdata.pl chief, predicts that the role of information verification specialists will strengthen, and new professions will be created that will, for example, verify AI output. Uhlig points out that journalists must above all be ready to work in extreme conditions, having in mind sudden events that changed the world, such as the pandemic and the Russian war against Ukraine. Therefore, future professionals should also learn how to talk to people, recognise falsehoods and extract the most important information from reliable sources.

During the online discussion on AI with Sobczak and Uhlig, Sebastian Ogórek, the chief editor of Wyborcza.biz, expressed concerns that AI will tailor its messages for concrete people by using psychological and social engineering tricks. As possible countermeasures in the future, he named new AI tools that will verify whether a given piece of information was generated by AI and state mechanisms that will not block the development of technology but will require the visible marking that the given information was created by AI.

“Our advantage over artificial intelligence is that we have ethics and that we still follow it in spite of everything. Not all, not always, but still”, Ogórek said.

It is certainly too early to say unequivocally how AI will affect the media market and the world. Newsrooms first need to learn how to use AI effectively and how to cover it. Refusing to use AI now is the same as quitting search engines 30 years ago, Sobczak told The Fix. Technologies change very dynamically and you always have to test new solutions; a different attitude would result in backwardness, she sums up.

Source of the cover photo: Adrian Grycuk via Wikimedia Commons

---