The UPF Morocco (French-speaking Press Union), in its training cycle, invited media consultant Gérarld Holubowicz to debate the issue last Saturday. The man, who has been teaching for ten years at Sciences Po in particular and the ESJ, deciphered, for members of the UPF, the genesis and the contours of an artificial intelligence which calls into question ancient customs to find information… In relation to the profession of journalism, it has a certain impact. But it should be taken with a grain of salt and used wisely. It’s a question of ethics! Lighting.
“This training is a proposal, a moment of discovery, it is not only a question of meeting your immediate needs but of acculturating yourself to generative AI to equip you effectively and allow you to better understand the future changes that are coming. are already announcing. From the outset, the expert will set the scene for his presentation so that the speakers situate themselves in this vast and heterogeneous world that generative AI represents. The latter can be defined as a recent phenomenon, which evolves at the speed of light, in terms of practices, the ethical and legal context, as well as on the industrial level. From there, Gérald Holubowicz already states that “it is therefore impossible to be exhaustive and address all the subjects of generative AI in one day of training”.
Everyone will then have to refine their knowledge in the matter according to their expectations. Industrial players limit free access to platforms, which makes the exercise less easy. In relation to the profession of journalist and in the audiovisual sector in particular, learning to write a prompt does not mean that the person knows how to use generative AI in its entirety. This broad domain implies that the person presents the context of the emergence of generative AI. It will also be about understanding the issues associated with generative AI. It will have to define the main use cases currently encountered in editorial offices. Because it’s a fact. Better understanding what generative AI can change in an editorial department is essential.
This makes it possible to identify the most responsible and ethical practices in an editorial production context. And this is why the training consisted in the second part of letting the journalists familiarize themselves and manipulate ChatGPT 3.5, 4.0, Perplexity.ai to understand the structure of a prompt. At the same time, the trainer continued to provide general information on the Chatbots to be tested. Several platforms were tested during the training, namely Perplexity.ai chatgpt claude and Gemini Copilot. The exercise also consisted of knowing how to produce content automatically… The partnership between Le Monde and Prisa Media is quite emblematic of such a practice. This does not mean that the journalist will be replaced but that he can be helped with his documentary research allowing him to construct his article in a factual manner with statements from people who have actually been interviewed to cross-check the information covered. The nuance is there. The question of ethics too. This is the very essence of the profession!
AI experts raise awareness about misinformation and ethics issues related to synthetic media. Gérard Holubowicz too.
With hindsight, everyone who has looked in depth at AI and its virtues raises awareness of its limits.
In the June 1, 2023 edition of Le Figaro, digital scientist Aurélie Jean rightly asserts that “contrary to what the creator of ChatGPT claims, an algorithm does not think.” For her part, the computer scientist specializing in the ethics of artificial intelligence, Timnit Gebru, director of DAIR and former Google ethicist, puts forward even more alarming remarks: “It is not surprising that the field of AI is is oriented in a direction that promises an unimaginable future on the horizon while it proliferates products that harm marginalized groups in the immediate future.”
As for the philosopher and author Eric Sadin, for him “AIs do not only speak for us, but in our name”. We will have understood that whatever the level of knowledge in AI, disinformation lurks. The trainer will focus on giving concrete examples on deepfakes. These materialize when a person takes someone’s image and makes them say something they didn’t say. This process is widely used to damage the image of politicians or bias judgment…
This is the case of the American campaign which ignited the web when Donald Trump appeared in a series of photos apparently taken with members of the black American community. These are fakes. In March 2018, American actor Jordan Peele called out the dangers that deepfakes and generative AI can represent…
Concerning synthetic media and translation, in 1947 Norbert Wiener planned to create a universal machine translation program to ensure “world peace”. In 2023, the HeyGen application realizes its dream.
In short, the profession of journalist should not disappear with generative AI. He must, however, be up to date with the latest advances in the field to know how to protect his writings and make controlled use of AI. That’s the whole point. In other areas as well. The health sector is also very coveted… To be continued.
AI challenges
Verbatim American computer scientist Fei Fei Li recalls the brilliant phrase, written in the 70s: “Today’s definition of AI is a machine that can make a perfect chess move while the piece is on fire” . This sentence perfectly illustrates the limits of AI. “In the next wave of AI research, if we want to create more useful machines, we need to return to understanding context,” she concludes.
The definition of Jean-Louis Laurière in 1986: “AI begins where classical computing stops: any problem for which there is no known or reasonable algorithm allowing it to be solved a priori falls under the AI.”
The definition of Yann Le Cunen 2017 “Artificial intelligence aims to allow machines to do tasks that are generally attributed to animals and humans.”Verbatim The American computer scientist Fei Fei Li recalls the brilliant sentence, written in the 70s: “Today’s definition of AI is a machine that can make a perfect chess move while the piece is on fire.” This sentence perfectly illustrates the limits of AI. “In the next wave of AI research, if we want to create more useful machines, we need to return to understanding context,” she concludes.
The definition of Jean-Louis Laurière in 1986: “AI begins where classical computing stops: any problem for which there is no known or reasonable algorithm allowing it to be solved a priori falls under the AI.”
The definition of Yann Le Cunen 2017 “Artificial intelligence aims to allow machines to do tasks that are generally assigned to animals and humans.”
