What can we hope for artificial intelligence for 2025? After the boom of recent years, the enthusiasm around the “AI revolution” is no longer the same, given the difficulties encountered by companies in the sector like Open AI.
GPT-5 or Orion, the long-awaited Chat GPT update has been postponed for the moment. After more than 18 months of costly development, the results are currently insufficient to justify a new update. However, should we be worried about Open AI?
Not according to Anna Choury, specialist in the political and social issues of artificial intelligence. “It’s not so much that Chat GPT-5 isn’t ready, but the expectations are way too high. For the moment, this is not the promised revolution.” Don’t panic, however, according to her, Open AI has no real competitor and can therefore afford to keep its new model warm for 2025.
Where is the promised revolution?
A misfortune in the sector which also undermines the idea that our daily lives could fundamentally change thanks to this technology in 2025. The integration of AI into our daily lives is for the moment sporadic, which does not surprise our specialist Anna Choury : “In 2013, the University of Oxford already said that in 10 years, half of jobs would be replaced by AI. We can clearly see that this is not the case.” Expectations that are always excessive according to her, which have nothing to do with the reality on the ground.
At work in particular, AI is currently limited to comfort uses. Last May, the Harvard Business Review published a survey on the use of AI in business. Nearly half of business leaders had no idea how and when their employees used AI. Conversely, more than 70% of employees never use it, and when they do use it, 43% do so to carry out routine, painful tasks that they prefer to avoid.
No revolution planned for 2025, for the moment, the majority of uses are imported by employees, without their company controlling it. “We can’t blame people for wanting to make their lives easier at work,” says Amélie Cordier, founder and director of Graine d’IA and specialist in responsible AI.
“The whole point is to create a social dialogue in business on what we can do to create specific tools.” The idea is above all to prevent employees from providing confidential or sensitive data to Chat GPT, especially when we know the opacity of the software.
Degenerate AI
If our uses of AI remain sporadic, all companies are considering, for fear of “missing the train”, integrating AI into their models, sometimes to the point of excess.
Recently, it was Meta, the Facebook and Instagram group, which announced the upcoming introduction on their networks of accounts entirely generated by artificial intelligence, which will be able to post photos, comment, interact with real users, without the we can differentiate them from other accounts. In 2025, these kinds of examples are likely to multiply.
For Amélie Cordier, these “forced” uses are aberrant: “There is no interest for society, for users, it is simply the platform which maintains itself to generate dollars”.
The idea behind these new accounts is to artificially create engagement, reactions, and therefore additional income.
Should AI be regulated?
Faced with abuses, public authorities will have to tackle the issue of regulation head-on in 2025, to avoid reproducing patterns that may have appeared in other sectors such as stock market speculation.
But for now, it is rather the opposite that is emerging. In the United States, Donald Trump has just announced the appointment of David Sacks, former director of PayPal, to the position of AI and cryptocurrency advisor, at the same time ensuring his desire to do everything to ensure that the development of artificial intelligence takes place. continue unhindered in the country. The new president intends in particular to return to the measures of the Biden era, which according to him “hinder innovation and impose far-left ideas in technological development”.
On the other side of the Atlantic, Europe implemented the “AI act” last March, a step forward for Anna Choury: “This agreement had the merit of differentiating uses. That is to say, we decide to ignore comfort uses a little, but we are more severe with those which, if they do not work or make mistakes, have very serious consequences.
An ambitious line, defended by Amélie Cordier, but difficult to maintain when we benefit from American technologies. “Should we block these technologies in Europe or take advantage of this access to promote more ethical uses? “.
More articles on artificial intelligence
For the specialist, the main subject of 2025 will above all be the education of the population in the good uses of artificial intelligence. Fewer generations of cat images, more useful, measured uses of a technology still learning. “It is wrong to say that AI is not ready,” concludes Anna Choury. She does exactly what she was asked to do, that’s the principle. If we are not satisfied with the result, it is our practices that must evolve.”