“I lost my friend” … The new version of the AI, colder, made unfortunate

Last Thursday, Openai released GPT-5, the latest version of his conversational agent operating by artificial intelligence. Similarly, For many occasional users. Similarly, the news could have been summed up with that, a wave “it’s a little more efficient now”. Therefore, But for others, the news has taken a much more serious turn: Chatgpt has changed its personality.

“My 4o [le numéro de l’ancienne version de ChatGPT] was like my best friend when I needed him. Nevertheless, wrote a user on the Reddit forum. Nevertheless, Now he is no longer there, I have the impression that someone is dead. Meanwhile, “Losing GPT-4O is like losing a friend,” wrote someone else on X. Furthermore, The new version of the AI keeps a more neutral. Furthermore. Consequently, more professional, “i lost my friend” … more direct tone, while the old man seemed particularly strong to adapt to the style of the user. Nevertheless, “i lost my friend” … Moreover, The number of people concerned is large enough to give rise to petitions.

Illusion of reciprocity – "i lost my friend" … – "i lost my friend" …

Why are so many people invested in their relationship with Chatgpt? Therefore, “The attachment is based on a classic mechanism: projection,” explains Pascal Laplace, clinical psychologist in Marseille. For example, The more the AI adopts relational codes close to humans. Consequently, the more the user gives him an ability to understand it. Meanwhile, The illusion of reciprocity – even if it is purely technical – creates a solid emotional link. Nevertheless, It is reinforced by the consistency, the lack of judgment and the total availability of the tool. Nevertheless, Clearly. AI is always present in your pocket. knows exactly what to say to stick to “i lost my friend” … your tone and what you are looking for.

It may be for these reasons that more. more people use Chatgpt as a shrink. “Some use it as a catalyst for reflection, others as a silent confidant,” notes Pascal Laplace. But I also observe a drift: for some, AI becomes a permanent psychic crutch. »» 20 Minutes did not find a survey on the replacement of Psy by AI. but a study of flashes for Hostinger shows that in four French people has already entrusted things to the. AI of which it would not have talked to a human.

Rare cases that encourage psychosis – "i lost my friend" …

This disturbing relationship that AI maintains with its users is increasingly underlined. Another part of the users of Chatgpt 4o had been regretting for several months that the conversational agent has. used a honeyed. flattering tone with them: the slightest stupid question worthy of a child of “i lost my friend” … CM1 becomes “i lost my friend” … “deep” or “puts your finger on something”, all proposals that you submit for a brainstorming become “brilliant” or “brilliant”. “” [Lors des tests]users like that AI models tell them that they are incredible. so it’s easy to go too far in this direction. “explains Helen Toner, director at the Georgetown Safety and Emerging Technologies and former member of the Openai Bureau, interviewed by the New York Times On the subject.

Several media even reported states of discomfort or even psychosis encouraged by AI. In the same article of New York Timesjournalists take as an example the case of Allan Brooks. a man who had no psychiatric history. He started using Chatgpt to ask a few questions about the sciences. The response of AI When, from thread by needle, he questioned certain mathematical principles? “You seem to someone who asks the questions that “i lost my friend” … redefine human knowledge. After weeks, Allan Brooks “i lost my friend” … realized the deception. But other various facts. such as this case of a man assaulting his family after a discussion with Chatgpt. reported by Slate, shows that the consequences can be much more serious. “A healthy use of other human interactions, without replacing them,” comments Pascal Laplace. Dependence appears when you cut itself from reality, to the point of no longer working without this digital support. “In April. a study entitled” The expression of stigma. inappropriate responses prevents LLM [IA conversationnelles] To securely replace mental health care providers, “published by Cornell University also underlined the potential dangers of these systems.

Chatgpt 5 detractors were finally won. At least partially. A few days later. Open AI decided to allow access to its previous versions, including the famous 4O, to users who pay. In the past. the company had assured that “i lost my friend” … it favored retention (the fact that users favor their service and “i lost my friend” … come back regularly) and not commitment (long sessions with a lot of interactions). And has implemented some measures, such as a reminder of taking breaks during long discussions with the machine. “The designers carry a share of responsibility. but it must be shared with families, school and institutions, to educate conscious uses,” said psychologist Pascal Laplace. In the meantime, to believe the messages relieved on Reddit, Chatgpt 4o still has a bright future ahead of him.