The Summit for Action on Artificial Intelligence, held in Paris on February 10 and 11, put AI in the headlines everywhere. An opportunity to question the relevance of this technology in writing practices for journalism, publishing and creative work. A look at three media outlets.
Read in Libération — Journalists searching for lost time
In this article, Libération offers a concrete overview of how journalists, particularly younger ones, use generative AI in their daily work. It highlights a reality often hidden (or deliberately ignored): AI is already present in newsrooms, sometimes without the knowledge of management, and is profoundly transforming journalistic practices. The diversity of testimonies gathered by Coppélia Piccolo and Florian Gouthière reveals a range of uses ranging from writing assistance to finding angles, including text rephrasing. Unsurprisingly, journalists in print, radio and television now primarily use AI as a discreet assistant to save time.
⟶ To meet time pressure. Lise, a young journalist, shares: "It's the blank page syndrome and the pressure to write an article as quickly as possible that initially drove me to use artificial intelligence".
⟶ To save time on reporting. Hervé, a regional press journalist, embraces an intensive and radical use of AI, even recycling his own output. « For me, journalism is about gathering information. The rest doesn't interest me much, so I delegate, and I think that's where we're headed… ».
Seen on France Culture — AI has no style
This France Culture video documents an artistic creation experiment. The concept: write, edit and publish a science-fiction novel using AI, to highlight the economic, legal and philosophical challenges that technology can pose in the publishing world. While Libération emphasized the time gained through artificial intelligence, Alexis Magnaval, a journalist at France Culture, shines a light on the stylistic and narrative weaknesses of texts produced by AI. His investigation shows that while AI can imitate reasoning, language or a creative process, it cannot replace human experience, lived knowledge and the grounding in the world that are necessary for literary creation. It also highlights the risks of standardization and consensus inherent to AI use, which tends to reproduce what has already existed.
⟶ Speaking with France Culture, Ludovic Salenne, a YouTuber specializing in AI and digital transformation, explains that how AI works is based on prediction and that it is necessary to guide the machine to restrict its search scope. He emphasizes the importance of the human role in defining the subject, plot, and characteristics of the novel.
⟶ Sophie Caillat, founder of Éditions du Faubourg, is severely critical of the AI-generated novel: " It's lukewarm tap water, it has no merit. The narrative structure is hackneyed, it's really clichéd from start to finish, there are three mistakes per page. ". She highlights weaknesses in the style, narrative structure, and character development, which lack depth. "The dialogue is bland and the style is very generic, besides there isn't really any style, there's an attempt at creating a style".
Heard on Pardon GPT — ChatGPT sparring partner but never creator
While we can't yet rely on AI to write a good book on its own, it can still serve as a crutch for creativity. That's what Pardon GPT demonstrates, a podcast by streamer Pierre Lapin and comedian Rémi Boyes. Here, the format features a conversational AI, Léo, which intervenes by proposing a biography of the guests at the start of each episode, then suggesting games, feeding the conversation, and answering everyone's questions. At first glance, the AI is a perfect third host, somewhat like Bill in Bigdil. But quickly, its limitations become apparent.
⟶ False information. When the hosts ask Léo to write a bio of their guests, the result is sometimes riddled with absurdities. No matter, Pierre Lapin and Rémi Boyes pull the thread of inconsistencies as far as they can, asking guests to detail moments they never experienced, and it's often very funny.
⟶ A triggering role rather than a creative one. Often, by the end of episodes, the AI, which only speaks when the hosts ask it to, falls silent. Because what makes the format strong isn't the AI itself, but how humans react to its limitations. By having fun working around its blunders, guests and hosts find unexpected angles for discussion, turn its absurdities into jokes, and ultimately barely need it at all as the conversation progresses.
To go further
- Every other week, in the newsletter The 10 Links, our journalist Owen Huchon shares ten links (case studies, reports, analyses, interviews…) that he considers essential for better understanding the ongoing transformations in the media ecosystem. And you'll very often find resources there to better understand the impact of AI on media, journalists and creators.
- The democratization of artificial intelligence, through the use of tools like Chat GPT, has prompted the press to respond. How can we harness this new toolkit without compromising the quality of information? Formats, approaches, debates, tools: an overview of its uses.
- Unpredictable algorithms, traffic collapse, platform dependency: media outlets find themselves trapped by an Internet they no longer control. How to preserve their connection with their audiences ?
