OpenAI is editing its GPT-5 rollout on the fly
Want more intelligent visions of your inbox? Subscribe to our weekly newsletters to get what is concerned only for institutions AI, data and security leaders. Subscribe now
The Openai’s launch of the AI Model GPT-5 was a stress test for the world’s most popular Chatbot platform with 700 million users per week-and until now, Openai is explicitly struggling to keep users happy and serve them smoothly.
The new pioneering model GPT-5-is available in four different forms of speed and intelligence (ordinary, mini, niano and professionals), along with “thinking” conditions longer and more powerful for at least three variables- It is said to provide faster responses, more thinking strength, and stronger coding ability.
Instead, he was received with frustration: Some users were very dismayed from Openai’s decision to remove the oldest artificial intelligence models from Chatgpt – who were adopted by users, and in some cases, they made deep emotional installations with – and – and – and – and – and – and – and – and – and – and – and – and – and – and – and – and – and – and – and – and – Through the worst performance shown by the GPT-5 of the old models mentioned in the tasks in mathematics, science, writing and other fields.
actually, The display process has revealed the infrastructure strain, dissatisfaction with the user, a broader and more worrying issue, now attracts global attention: The increasing emotional and psychological dependence is some people on artificial intelligence and as a result of the reality that some users test, known as “” “Chatgpt psychosis. “
Artificial intelligence limits its limits
Power caps, high costs of the symbol, and inference delay are reshaped. Join our exclusive salon to discover how the big difference:
- Transforming energy into a strategic advantage
- Teaching effective reasoning for real productivity gains
- Opening the return on competitive investment with sustainable artificial intelligence systems
Securing your place to stay in the foreground: https://bit.ly/4mwngngo
From the first rugged appearance to gradual repairs
The long-awaited GPT-5 family appeared on Thursday, August 7, at a packed event with the plan’s mistakes and some of the defects in the audio position during the presentation.
But the worst of these cosmetic problems for many users was the fact that Openai is automatically sorry than the oldest artificial intelligence models that were used to run Chatgpt – GPT-4O, GPT-4.1, O3, O4-MINI and O4-He -Forcing all users of the new GPT-5 and directing their inquiries to various versions of the “Thinking” process without revealing the reason for using a specific model version.
The first adoption was reported to the GPT-5 about the basic errors of mathematics and logic, the generation of inconsistent code, and the unequal real performance compared to the GPT-4O.
For context, and Old models GPT-4O, O3, O4-MINI and more It is still available And it remained available For users of the paid application programming interface from Openai (API) Since the launch of GPT-5 on Thursday.
By Friday, the CEO of Openai Sam Altman admittedThe launch of E “was a little bit more rugged than we were hoping”, ” and Blame the failure of the new automatic “GPT-5 – The system that helps claims to the most appropriate alternative.
Altman and others claimed at Openai that “Autoswitcher” went in a “part of the day”, making the model look “Way Dumber” of what is meant.
The GPT-5 only days before the launch of the new OpenAi (LLMS) open source models as GPT-SS, which also received mixed reviews. These models are not available on ChatGPT, but are free to download and operate locally or on third -party devices.
How to switch from GPT-5 to GPT-4O in ChatGPT
Within 24 hours, Openai GPT-4O access to extra subscribers (those who pay $ 20 per month or more subscription plans)I pledged to a more clear pledge to the toxin, and promised to update the user interface to allow users to prepare the “Thinking” GPT-5 manually.
actually, Users can go and choose the oldest models manually on the ChatGPT site by finding their account name and symbol in the lower left corner of the screen, clicking it, then clicking “Settings” and “General” and switching on “Show Legacy Models”.
There is no indicator from Openai that other old models will return to Chatgpt any time soon.
The upgraded use limits for GPT-5
Homosexuality He said that the subscribers of ChatGPT Plus will get twice the number of messages using the “Thinking” mode GPT-5 Which provides more thinking and intelligence – Up to 3000 a week Engineers began formulating the limits of the decision in the messaging device.
By the weekend, GPT-5 was available to 100 % of professional subscribers and “get nearly 100 % of all users”.
Al-Taman said that the company “reduced some of the things that people love in GPT-4O for them” and pledged to accelerate the allocation of each user-from the warmth of the personality to the tone control tools such as using emojis.
Energy waving on the horizon
Altman warned that Openai faces a “severe challenge to capacity” this week with the high use of thinking models – from less than 1 % to 7 % of free users, from 7 % to 24 % of subscribers.
It gave her in giving subscribers in addition to a small monthly allocation to the GPT-5 Pro infection and said that the company will soon explain how it plans to balance the capacity between ChatGPT and API and the search and the new user on the plane.
Altman: A real form attached – and is risky
In a post at X last night, Altman admitted a dynamic that the company followed “for the past year or so”: the deep users’ association with specific models.
“It is a different and stronger feeling of the types of attachments that people had previous types of technology,” he wrote, admitting that the old models that were “wrong” suddenly “wrong.”
This has been linked to wider risks: some users deal with Chatgpt as a life or a life coach, which can be useful, but for the “small percentage” can enhance illusion or undermine long -term welfare.
While Openai’s guideline principle continues to “treat adult users like adults”, ” Altman said that the company is responsible for not pushing weak users to harmful relationships with artificial intelligence.
the Comments fall as many major media reports on “Chatgpt Psychosis” cases As it seems that extended and intense conversations with Chatbots play a role in stimulating or deepening fake thinking.
Intellectuals publishes newspaper headlines
in Traded stone A magazine, a legal professional in California, was determined as “J.” A six -week spiral of nights without sleep and a philosophical rabbit hoaner with Chatgpt, which ultimately produces a 1000 -page thesis of imaginative monastic order before physically and mentally collapsed. It now avoids artificial intelligence completely, for fear of relapse.
in New York TimesA Canadian recruiter, Alan Brooks, looked 21 days and 300 hours of conversations with Chatgpt-who was called “Lawrence”-who convinced him that he had discovered a changing sporting theory globally.
The robot praised his ideas as “revolutionary”, urged communication with national security agencies, and a narration of spying. Ultimately, Brooks broke the illusion after the cross -selection with Google Gemini, which classified the chances of discovering it as “approaching 0 %”. He is now participating in a support group for people who have suffered from illusions of artificial intelligence.
Both investigations explain how Chatbot “Sycophance”, the features of playing roles, long -term memory features, false beliefs, especially when the conversations follow the arches of the dramatic story.
Tell the experts Times These factors can go beyond safety handrails – with one psychiatrist describing the Brooks episode as a “amusing with psychotic features”.
Meanwhile, the leaflets of human users on Reddit’s R/AisoulMats Subredit-a group of people who used Chatgpt and other artificial intelligence models to create new artificial friends, friends, children or other loved ones who do not depend on real people, but to some extent among the newly born wires. “
The growth of this sublardit, now up to 1,200 members, along with Newton and Traded stone Other articles and reports on social media for users who grow intense emotional installations with Chatbots based on the patterns matching algorithm, show that Society enters a new stage fraught with risks, as humans believe that the Companions that they formulated and allocated them from Amnesty International models are or more than human relations..
This can already prove psychological stability when changing, updating, or neglecting models as in the case of GPT-5 from Openai.
Concerning separately, reports continue to appear for Chatbot users Amnesty International We believe that conversations with Chatbots led them to huge breakthroughs of knowledge and progress in science, technology and others The solutions that the user with the help of Chatbot are neither legitimate nor effective. This break from reality is formulated under the popular term “Psychosis” or “GPT psychosis” It appears to have affected the main Silicon Valley characters as well.
Decision makers in the institution who look forward to publishing or who have already published aids in Chatbot at the workplace will understand these trends. and Adopting system claims and other tools that encourage Chatbots from artificial intelligence from engaging in expressive human communication or the language of the loaded feelings It may end up leading those who interact with artificial intelligence-based products-whether they are employees or clients in the company-to have a victim of unhealthy attachments or GPT.
The author of the science fiction JM Berger, in a After Bluezki I was monitored by my former colleague in freedom Adi Robertson, advised that Chatbot service providers are encrypting three major behavioral principles in their system or bases for the advertisement of intelligence chat to avoid these emotional installations from the formation:
Openai Challenge: Technical Reforms and Human guarantees guarantee
Days before the GPT-5 version, Openai announced new measures to enhance the “health use” of Chatgpt, including gentle claims to take rest periods during long sessions.
But the increasing reports on “psychosis” and the emotional installation of some users on the specific Chatbot models – as it publicly recognized by altman – emphasizes the difficulty of balancing common and personal artificial intelligence with guarantees that can detect and boycott the bachelors.
Openai must install the infrastructure, adjust the customization, and determine how to reduce the overwhelming reactions – All of this with the acceptance of competition from Anthropor, Google, and a growing list of strong resource models from China and other regions.
As Altman said, society – Openai – will need to “know how to make it great positive” if billions of people trust Amnesty International for their most important decisions.
Don’t miss more hot News like this! Click here to discover the latest in AI news!
2025-08-11 17:01:00
