“I give the floor to the accused! » says the president in a solemn tone. And suddenly, above the members of the court, the image of a dark-haired woman in her twenties appears on the big screen. “Honorable members of the court,” the creature attacks in a smooth voice with slightly mechanical inflections, “I stand before you as Anne Kerdi, a voice embodying artificial intelligence in this crucial trial. »No, pleads the creature, in a rapid flow, articulating each word, she is “in no way a threat to employment, nor to all of humanity”. No, ladies and gentlemen of the jury, it will not “supplant humans”, but “lighten their burden by automating repetitive and tedious tasks”.
Dumbing us down, her? On the contrary, it promises to “provide us with unprecedented access to knowledge and information,” and heralds a bright future “where artificial intelligence and humanity will blend harmoniously for the benefit of all.” Fine point, the accused gets a hit: can we judge her, when after all, she is “neither endowed with her own conscience nor with an autonomous will”, but is only a “computer program” that “acts within the guidelines and parameters established by [its] designers”?
We will no longer see or hear the accused during the nearly three hours that her fictitious trial lasted, organized on Thursday September 14 by the Noûs think tank in a municipal hall in Montpellier, in the presence of 300 spectators. But a host of experts, witnesses for the prosecution or the defense, of artificial intelligence, who had to answer for the worst crimes: “threats to millions of jobs around the world”, “dissemination of false information and false images ”, and for that matter, “endangering humanity.”
More than the evanescent Anne Kerdi, a fictitious Instagrammer created by a Breton developer to promote her region, it was the generative AI ChatGPT which quickly found itself in the dock. “How can we assess our students’ ability to reason, their overall understanding, if they have used an AI? » asks Liliane Dorveaux, deputy director of the EPF engineering school. “I regularly use ChatGPT to program, write texts, cross-reference my sources to have more reliable information,” testifies Sébastien Dors, young apprentice engineer, and president of a “Generation AI” working group. He himself, who “grew up without AI”, believes he was able to “develop his critical thinking”. But he worries about the next generations born after AI, “who will not have the same discernment.”
Digital assistant for the City of Montpellier, Manu Reynaud explains why his municipality has “banned ChatGPT to town hall agents for their professional use”, in the name of the “precautionary principle”. A “citizens’ convention” will work on the subject. The public enthusiasm for ChatGPT has “amazed” scientists, recognizes Jean Sallantin, a Montpellier researcher considered one of the fathers of AI in France. Even though it is, he says, “old technology” which has nothing revolutionary in itself.
In defense, Grégoire Mercier, head of the “Health Data Sciences” medical team at Montpellier University Hospital, touts the benefits of AI in the medical field: it makes it possible to “make diagnoses in an extremely efficient way”, offers patients “access to cutting-edge medicine, wherever they are”, and participates in “the discovery of new treatments, new chemical molecules”.
But AI, recalls jurist Léah Perez, also presents numerous risks: “for the safety of people in critical sectors such as health or aeronautics”, for “our fundamental freedoms and rights” with “facial recognition for security purposes.” And for our very lives: “There was recently a case of suicide in Belgium following a discussion between a person and a conversational agent”, a ChatBot. Faced with the absence of a general text governing AI, the European Union, she explains, is preparing a regulation to classify them according to risks.
Michel Robert, director of Cines and the Adastra supercomputer, also calls for “regulating AI”. And points out its “dark sides”: the use of energy-intensive technologies, and “underpaid labor in Madagascar to adapt the responses”. He does not believe in a “new civilizational turning point” with the advent of AI. Neither does Jean Sallantin: “The obvious revolution is that of the climate crisis. » And in his eyes, AI can help us overcome it, “by creating tools that are used to lead efficient debates” in the face of this crucial issue. At the end of the trial, the humans present are invited to decide. Unanimously, the AI ??is acquitted: unlike HAL, the paranoid computer from 2001: A Space Odyssey, it will not be disconnected. For the moment.
