After the excitement, it’s time for selection at the editorial offices

Simple support or tool for transformation: Many journalists are testing artificial intelligence (AI) and editorial staff are feeling the issues at hand, from hiring to ethics.

The question is on everyone’s mind at the International Journalism Festival in Italy, in Perugia, until Sunday.

Artificial intelligence tools, which mimic human intelligence, are commonly used in editorial offices around the world to transcribe audio files, summarize texts and translate… so much so that in early 2023 the Axel Springer Group in Germany announced job cuts at its daily newspapers Bild and Bild. Die Welt newspaper, on the grounds that artificial intelligence can now “replace” some journalists.

Over the past year and a half, generative AI, which makes it possible to obtain text and images based on a simple query in everyday language, has opened the way to new uses, raising new concerns. Voices and faces could, for example, be cloned to produce a podcast or a television news show. Last year, the Philippine site Rappler was able to create a brand targeting young people, by turning its long-form articles into comic stories, graphics, and even videos.

Media professionals agree that a career as a journalist will focus on higher value-added tasks. “You are the ones who do the work” and “the tools we produce are enablers,” Google News CEO Shailesh Prakash insisted in Perugia.

The cost of entry into generative AI has dropped with the arrival of ChatGPT in November 2022. The tool created by US startup OpenAI is now available for small newsrooms.

Inspired by this, and with the help of engineers, Colombian investigative media Cuestión Pública has developed its own tool that allows, in the event of “breaking news,” to search its archives for elements of context. “This can be edited instantly in our app,” confirms Claudia Báez, General Director of Cuestión Pública.

Toutefois, “The number of sociétés that press no products with their own language models”, at the base of the IA interfaces, observes Natali Helberger, professor at the University of Amsterdam, who satisfies the best technologies “sûres et dignes trust “.

EveryPixel Journal estimated last summer that there were as many AI-generated images in one year as were created in 150 years of photography. In the face of this tidal wave of artificial content, how can information emerge?

In the face of “deepfakes” (elaborate montages), media and technology actors are uniting, especially with the Content Provenance and Authenticity Alliance (C2PA), which seeks to define common standards.

“The core of our work remains information gathering and field reporting. We will be relying on human reporters for a long time to come,” recalls Sophie Hewitt, recently appointed deputy CIO responsible for artificial intelligence at AFP.

Reporters Without Borders (RSF), which has expanded its work to defend reliable information, launched the Paris Charter on AI and Journalism at the end of 2023. “It emphasizes transparency. How much will publishers have to disclose when they use generative AI?” », asks Anya Shiffrin, Director of Training at Columbia University (USA). The debate continues at the Swedish public broadcaster: “Should all content be identified or should users trust the brand?”, reports Artificial Intelligence Director Ole Zachrisson.

Regulation is still in its infancy in the face of ever-evolving technology. The leading European Parliament adopted a text in mid-March that regulates artificial intelligence models, without restricting creativity. As for the editorial board itself, charters or guidelines for good practices are widespread. “We change our guidelines every three months,” says Ritu Kapoor, head of Quintillion Media in India. No article can be written by AI and the images generated cannot represent reality.

AI models need to be fed by data. Without paying suppliers? In December, The New York Times filed a lawsuit against OpenAI, the company that created ChatGPT, and Microsoft, its main investor, for copyright infringement. In contrast, other newspaper groups have signed agreements with OpenAI, such as the German company Axel Springer, the American press agency AP, and more recently the French daily newspaper Le Monde, as well as the Spanish group Prisa Media (El País, As).

While journalism lacks resources, it’s tempting to collaborate, says Emily Bell, a professor at Columbia University’s journalism school. She senses some form of external pressure: “Get on the plane, don’t miss the train!” »