OpenAI Scrambles to Update GPT-5 After Users Revolt

OpenaiThe GPT-5 model was supposed to be a global variable upgrade to the famous and innovative Chatbot. But for some users, last Thursday’s version seemed to reduce the amazing classification, as the new Chatgpt presents a surprisingly stupid character and stupid lines.

On Friday, Sam Altman, CEO of Openai, took to X to say that the company will keep the previous model, GPT-4O, which works for excessive users. Altman said that a new feature is designed to switch smoothly between models depending on the complexity of the inquiry was broken on Thursday, “The result GPT-5 seemed stupid.” Promise to implement repairs to improve GPT-5 performance and the total user experience.

Looking at the noise around the GPT-5, it seems that there is an inevitable level of hope. When Openai GPT-4 presented in March 2023, artificial intelligence experts surprised his amazing capabilities. GPT-5, the concerns, will definitely be like dropping the jaw.

Openai described the model as an important upgrade, with doctoral intelligence skills and Virtuoso coding skills. A system to automatically aim to direct the various models to provide a smoother user experience. (The company can also provide the company’s money by directing simple inquiries to cheaper models.)

Shortly after the GPT-5, however, the Reddit community is intended for chatting full of complaints. Mounds of many users to lose the old model.

“I have been trying to GPT5 for a few days now. Even after allocating the instructions, he still doesn’t feel the same. It is more technical and more generalized, and he feels honestly emotion,” a member of the community wrote in a thread entitled “Kill 4O is not innovation, it is erased.”

“Certainly, 5 on what you are – if you hate differences and feel things,” another user wrote.

Other topics of slow responses, hallucinations and sudden errors complained.

Altman promised to address these problems by doubling the GPT-5 rate of Chatgpt Plus users, improving the system that turns between models, and allows users specifically when they want to run the most capable and capable “thinking mode”. “We will continue to work to get things stable and we will continue to listen to the comments,” the CEO wrote. As we mentioned, we expected some collision because we put a lot of things at one time. But it was a little more rugged than we were hoping! “

The mistakes published on social media do not necessarily indicate that the new model is less capable of its predecessors. They may simply suggest that the all -new model is stumbled in cases different from previous versions. Openai exactly the comment on the reason for the appearance of GPT-5 sometimes to manufacture simple errors.

The reverse reaction sparked a new discussion about the psychological attachments that some users are formed using Chatbots trained to push their emotional buttons. Some Reddit users rejected complaints about GPT-5 as evidence of unhealthy dependence on artificial intelligence companion.

In March, Openai published a research exploring Bonds emotional users with its models. Soon after, the company released an update to the GPT-4O after it became very very surcophanty.

“It seems that the GPT-5 is less sound, more” business “and less shred.” “I personally think this is a good thing, because it also led to delusions, promoting bias, etc., but unfortunately, many users love a model that tells them that they are smart and amazing and confirm their opinions and beliefs, even if [they are] mistake.”

Altman indicated in another post on X that this is something that the company struggles in building GPT-5.

Al -Taman wrote: “A lot of people actually use Chatgpt as a type of processor or life trainer, even if they do not describe it in this way,” Al -Taman wrote. He added that some users may use Chatgpt in ways that help improve their lives, while others may be “lifting unaware of their long -term well -being.”