After making headlines for weeks in Quebec, the artificial intelligence tool ChatGPT is timidly entering newsrooms here. Considered a promising avenue to facilitate certain tasks for journalists, could the tool in the long term completely replace them? Overview.
“NB: This text was written with the help of artificial intelligence (AI) from the QUB radio audio interview. Since the beginning of May, this mention has appeared at the end of several articles on the QUB radio website.
Without much fanfare, Quebecor’s digital radio began using the popular chatbot of OpenAI which is based on the latest human language understanding technology to generate consistent answers on various topics.
For the moment, at QUB, it mainly makes it possible to write short journalistic texts to accompany audio extracts taken from radio broadcasts, explains in an interview the general manager, Jean-Nicolas Gagné. “We take the content of the interview to transcribe it and we ask [à ChatGPT] to provide us with a text of a hundred words. There is always someone who proofreads everything, who “fact-checks” and rewrites a little if necessary”, he specifies.
Each time the application is requested, the media makes a point of specifying it to its readers, for the sake of transparency.
“We are really exploring, there is really no rush to go faster to integrate it into the newsroom,” continues Jean-Nicolas Gagné.
” Caution “
The other major media contacted have not yet published an AI-generated article. On the other hand, they are currently analyzing the possibilities in this direction, while calling for caution.
“If AI is better than humans at performing tasks that have no added value, we will not refrain from considering its use”, says the director of the Duty, Brian Myles. The daily is “in deep reflection” and “does small tests” internally, but “no clear strategy” has been adopted. “It’s premature to do it now because it’s just starting to emerge and it’s going very fast. We want to give ourselves the time to think carefully, ”he adds.
An opinion shared by François Cardinal, assistant editor at The Press. “We have launched a study committee on a possible ethical framework,” he says. But I confess to being skeptical when I see media doing like Wired with rules already established, while we are barely beginning to grasp the full potential and all the possible uses, which makes the rapid adoption of a guide risky. »
Radio-Canada’s director general of information, Luce Julien, maintains that “AI is essential” and that we must “learn to tame it”. “We are moving forward on these issues with caution and vigilance. »
A threat ?
Thus, the media are preparing and considering that ChatGPT will cause a small revolution in their newsroom, as the arrival of the Internet did decades ago.
Could ChatGPT and its counterparts even replace journalists? “I really don’t see that day coming,” drops Jean-Nicolas Gagné, of QUB radio, recalling that the use of chatbot requires the supervision of employees. Other media interviewed share his opinion, questioning the reliability of the tool.
Brian Myles, of Duty, criticizes in particular his “mythomaniac” side. “The application invents subjects, makes causal relationships where there are none, juxtaposes historical events. There are still a lot of flaws. Several articles published in the media recently have reported on this problem.
Journalism professor at UQAM Jean-Hugues Roy agrees and raises other flaws. Taking the example of AI-generated texts published by QUB, he notes a lack of precision, repetitive formulations and finds the choice of citations sometimes irrelevant. “A human would clearly do better,” he judges.
“The job of journalists is also to search through documents, to check the facts, to go into the field to meet people to give them a voice. It takes humans to do that,” he says. According to him, ChatGPT is therefore not a threat for journalists, but a tool that they must master to save time on daunting tasks.
This is already the case with other AI tools that have integrated most newsrooms in recent years: transcription of interviews, SEO research, publications on social networks and other exploratory projects.
“The most worrying thing about ChatGPT is not its use in the media. But the one that content firms or simple individuals will make of it, who risk spreading a lot of disinformation, ”concludes Jean-Hugues Roy.