How to keep control?
“Ladies and gentlemen, good evening!” The Cannes 2026 festival is proud to present to you Criterionthe first animated film made largely by an artificial intelligence (AI). Admire these nice monsters, with blue, pink or red fur, with big teeth and fluffy voice … “Next spring, a scene of this kind could perfectly take place during the big meeting of world cinema.
The American company Openai has indeed announced that it intended to present to Cannes its unprecedented creation whose images and animation have been entrusted to the machine. The film cost $ 30 million, three times less than big American productions. A new example of the “surge” of the AI.
Until then, everything was still relatively simple. We knew “classic” artificial intelligence, which uses computer programs to analyze data automatically. It irrigates digital systems – GPS, postal sorting program, search engine, algorithms – and has constant progress.
Used in medical imaging, it improves diagnosis and can even detect unknown primitive origin cancers, those who have metastasized without knowing which organ was hit first. It also actively contributes to discovering new drugs.
On a daily basis, it slips into the algorithms of recommendations from our social networks, in our car programs, in video surveillance cameras, etc.
But here we are with the Chatgpt conversational robot of Openai, which appeared in 2022, at the so -called “generative” AI. A major technological jump and a societal upheaval. This AI itself creates, on request, unpublished content from the data it collects on the web.
More precise than traditional search engines, this chatbot (conversational robot, in French) is available everywhere and at any time, can synthesize in a few seconds astronomical sums of knowledge and answer the least of our questions.
It helps to write texts, the synthesis of documents and even serves as psychological support for some … but with errors, too. THE Deus ex machina Modern times is not (yet) infallible.
An prolonged digital earthquake
Let’s get back. The first revolution – coal and steam machine – passed England from the 18th century from a rural country, dominated by agriculture and crafts, to a manufactured products factory. The second – electricity, oil and chemistry – accelerated globalization and economic flows between the 19th and 20th centuries. The third – fleas, screens and internet -, initiated in the 1970s, introduced digital into our lives.
Computer, smartphones and networks have transformed social and economic ties in depth. More than a fourth industrial revolution, we are experiencing the spectacular extension of this digital earthquake. “The idea of artificial intelligence is concomitant with the birth of computer science in 1945”, recalls Nicolas Sabouret, researcher in AI at Paris-Saclay University.
The term “artificial intelligence” also appeared in 1956, cross -checking “any program capable of performing tasks that we execute with our intelligence”. Advanceds are such that the latest GPT-5 language model still pushes the limits: in a study by the University of California in San Diego not yet published*, some participants considered it more human than real humans!
However, the machine does not converse. It only responds by offering a series of words according to a calculation of probabilities. For this, it is based on the gigantic quantities of data gleaned on the internet allowing it to create new data in turn: images, texts, sounds … AI therefore ends up feeding on content … generated by itself, thus reducing the quality of its productions.
Impacts and productivity
Is the generative explosion of AI as a host of questions: “Is the source of information provided?” Let us gradually start our capacity for reflection by delegating it to software? ”
The philosopher Anne Alobert, author of Artificial stupidity (Ed. Allia), notes, in fact, a break in “social knowledge”: “When we write a prompt (A question or request), we give an order to software that makes calculations, not to human otherness supposed to understand and interpret, she observes. The machine produces an automatic language which dispossesses us of our ability to express ourselves in writing and orally. We lose the singularity of expressions. »»
A study in the prepublication of the renowned Massachusetts Institute of Technology (MIT), very recently relayed, seems to corroborate this hypothesis. But in the very opinion of its authors, the panel – 54 people questioned – does not allow scientific conclusions. Research is only in their beginnings.
As for the improvement in productivity that AI is supposed to offer by eliminating repetitive tasks, the Nobel Prize in economics Daron Acemoglu estimates the growth and productivity gains at 0.7 % over ten years, and at 5 % the human tasks that will be replaced.
Like other technological revolutions, artificial intelligence could weaken certain trades, such as those in the creation sector. Sociologist Juan Sebastian Carbonell speaks of “augmented Taylorism” pre -leaving more than it increases the number of workers.
“Many technologies adopted are of very poor quality and have little productive effect. We are talking about a fashion effect, adds Malo Mofakhami, lecturer in economics at Sorbonne-Paris-Nord University. We are in a profusion phase with an amazement effect that pushes companies to adopt AI at full speed. But discomfort at work linked to these technologies is massive, adds the expert. The pace of change increases at full speed, but it takes time to train employees. ”
The human still in the race
The large -scale dismissal plans observed among tech giants, however, do not lead to a disappearance of work as well as its reconfiguration. “Sociology work shows that AI never works alone. You need people to check and formulate requests… ”, continues Malo Mofakhami. Human has not said their last word.
How to regain control, when AI is progressing by giant? “We absolutely have to think about an ecologically sustainable artificial intelligence – it is very energy -consuming – and socially interesting for the collective,” invites Anne Alobert.
Initiatives already exist. The Sunflower project located in Switzerland, for example. This collaborative platform, a sort of wikipedia against false information, was launched in 2021 by a team of researchers and engineers from various countries. Its users note and compare content according to several criteria: accuracy of the facts, scientific rigor … Then they issue recommendations.
Rather than enclosing Internet users in personalized informational bubbles, the platform thus highlights content deemed of general interest. What if the response to the dangers of the AI was in the collective and international cooperation?
