OpeniThe GPT-5 model meant to be a world-ups upgraded for its wildly popular and uncertain chatbott. But for some users, last Thursday’s release felt like a drenched downgrade, in which the new chat introduced a thin personality and made surprisingly dumb mistakes.
On Friday, Openai CEO Sam Altman took the X to say that the company is running for the previous model, GPT -4O for plus users. A new feature designed to switch between the models based on the model based on the complexity of Query was broken on Thursday, Altman said, “And the result was GPT -5, it was Dumber.” He promised to implement reforms to improve the performance of GPT-5 and overall user experience.
Given the campaign around GPT-5, some level of disappointment appears unavoidable. When Openai introduced the GPT-4 in March 2023, it surprised AI experts with its incredible abilities. GPT-5, the pundits estimated, certainly only as leaving the jaws.
Openai postponed the model with PhD-level intelligence and virtuous coding skills as an important upgradation. A system was to provide a smooth user experience for various models automatically root the query. (It can also save the company’s money by directing simple questions for cheap models.)
Immediately after the GPT-5, however, a Reddit community was dedicated to a chat filled with complaints. Many users mourned the loss of old models.
“I have been trying for GPT5 for a few days now. Even after customizing the instructions, it still does not feel the same. It is more technical, more generalized, and sincerely feels emotionally away,” “Kill 4o innovation, it’s not eradication.”
“Surely, 5 is fine – if you hate the nuances and feel things,” another Reddit is written by the user.
Other threads complained of sluggish reactions, hallucinations and stunning errors.
Altman promised to address these issues by doubling the GPT-5 rate limits for the Chatgpt Plus users, which improved the model improvement system, and allowed users to specify that when they want to trigger more thoughtful and competent “thinking mode”. “We will continue to work to stabilize things and keep listening to the response,” the CEO wrote on X.
The errors posted on social media are not necessarily less capable than its predecessors. They can only suggest that all new models are implicated in different edge cases compared to pre-versions. Openai specifically refused to comment on why GPT-5 sometimes appears to make a simple blunder.
Baikalash has provoked a new debate on psychological attachments, some users have created with a trained chatbot to push their emotional buttons. Some reddit users dismissed complaints about GPT-5 as evidence of an unhealthy dependence on AI partner.
In March, Openai published research searching as emotional bond users with its models. Shortly thereafter, the company issued an update to GPT-4O, as it was very smooth.
“It seems that GPT -5 is less smooth, more” business “and less spicy,” a professor of MIT, Patty Mess, who worked on studies. “I personally think of it as a good thing about it, because it is also what happened due to confusion, prejudice reinforcement, etc., but unfortunately many users prefer a model that tells them that they are smart and amazing and it confirms their opinion and faith, irrespective of even if [they are] Wrong.”
Altman indicated in another post on X that it is something that the company wrestled in the construction of GPT-5.
“Many people effectively use Chatgpt as a type of physician or life coach, even if they will not describe it in this way,” Altman wrote. He said that some users may use Chatgpt in ways that help improve their lives, while others “unknowingly may be away from their long periods.”
