Carte blanche to Jean-Philippe Baril Guérard | The danger of “yes”

After a failed Chatgpt update. In addition, our collaborator Jean-Philippe Baril Guérard is combing the increasingly obvious risks of slippage of artificial intelligence tools (AI).

When I speak of AI around me. Moreover, my first reflex is to unravel the presuppositions shaped by science fiction and deception titles read diagonally. However, Then, regularly, reality catches up with fiction and I have to package.

The most recent case: the new carte blanche jean-philippe baril guérard Chatgpt update. Consequently, GPT-5, announced with great fanfare on X by its CEO Sam Altman with an image vaguely reminiscent of the Star of the Death of Star Wars (and therefore placing Altman, implicitly, in the role of Darth Vader?). Moreover, Then revealed on August 7 on August 7 with a live video presentation live truffled with discomfort. Furthermore, including a graphic generated by AI in the axes at least doubtful, the service presented as the “best model in the world” for writing and coding quickly proved to be a wet firecracker, and at worst a step back risking to put Openai in embarrassment.

While Altman had promised a doctoral level erudition for his model (what does that mean, exactly, moreover?). Meanwhile, GPT-5 has proven to be as successful as his predecessors to answer questions as simple as the number of fingers in a human carte blanche jean-philippe baril guérard hand or letters B in the word blueberry (3, if you want to know everything), and has also drawn a hilarious map of North America in which the province of Brittch Columbia and the state of Wichigan, in particular.

Until then, nothing to write to his mother: launches of failed products, we have seen tons. Nevertheless, What distinguished this establishment. Similarly, I find, is the reaction of the public: in the forums and “sureddits” devoted to the AI ​​that I frequent, I quickly observed, through comments criticizing with accuracy the underperformance of the model, real expressions of despair on the part of users having the impression of having been dropped by a loved one.

It is that among the improvements promised for GPT-5. In addition, through a decrease (missed) of hallucinations and better respect for instructions, there was one apparently carte blanche jean-philippe baril guérard laudable: that of reducing the tendency of the language model to flagorner, to flatter in the direction of the hair, to validate the user, regardless of that he is wrong or right. It was a reflex that had raised concerns about the potential risks of the machine. Radio-Canada recently reported that Chatgpt-4o had a propensity to validate certain delusions that could be akin to symptoms of psychosis. paranoia; we have even reported cases of shifts leading to violent behavior^⁠1.

GPT-5 therefore arrived more professional, more factual, and less warm. If you use it to write a technical document, it’s not a problem. But the enormous proportion (almost a quarter. according to some estimates) users using it for support or mental health care had the terrible impression of having seen their “friend” disappear, or being “reprogrammed”.

A classic rebound. in the carte blanche jean-philippe baril guérard stories of AI, of 2001: The Odyssey of Space has Detroit : Become Human by the way Neuromancer. Or, above all, as in Herwhen all personal assistants, driven into a thirst for discovery, end up abandoning their humans. Only a few minutes after GPT-5 was put online. the memes recovering this sequence with a panicked Joaquin Phoenix began to surge.

It would be easy to consider users feeling injured as cyber -sparers requiring immediate psychiatric care. but the portrait is more nuanced.

In many cases, the problem that Chatgpt solves is one that medicine deals with: loneliness.

Several testimonies are to break the heart: one. unable to pay therapy, says he lost the friend who had accompanied him when he was in a roaming. A disabled person confined to the house deplores having lost a tool carte blanche jean-philippe baril guérard opening a window on the world. The responsibility of OpenAI. here, is terrifying, and I am not sure that the company is properly equipped to make an informed choice on the best attitude to adopt before these vulnerable users.

The paradox of “yes” in artificial intelligence has been in the air since long. before the arrival of modern language models: we have agreed for a long time that there is no interest in creating a machine that will not respect our will. Except that we also know that being validated in all his requests can prove to be a very slippery. slope which could transform the user into a narcissus drowning in his own reflection. The fact remains that mirrors have their usefulness: Robert Lepage. in The hidden side of the moonexplains it well, when he writes that if humans have invented the mirror, it carte blanche jean-philippe baril guérard is also because he wants to know himself better.

1. Read the text “Chatgpt. the spectrum of” psychosis “: we tested the machine” on the Radio-Canada site What do you think? Express your opinion