Friday, the CEO of Openi Sam Altman taken ax To say that the company would maintain the previous model, GPT-4O, running for more users. A new functionality designed to pass seamlessly between the models depending on the complexity of the query had broken on Thursday, Altman said: “and the result was GPT -5 seemed much more stupid”. He promised to implement corrections to improve GPT-5 performance and the overall user experience.
Given the hype around GPT-5, a certain level of disappointment seems inevitable. When Openi introduced GPT-4 in March 2023, he amazed artificial intelligence experts with his incredible skills. GPT-5, hypothesizes the experts, would certainly be equally amazing.
Openii has propagated the model as a significant update with the doctoral intelligence and virtuous coding skills. A system to automatically instract the query on different models had to provide a more fluid user experience (it could also save money by the company by directing simple questions to cheaper models).
Shortly after GPT-5, however, has fallen a Reddit community dedicated to Chatgpt full of complaints. Many users have cried the loss of the old model.
“I have been trying GPT5 for a few days. Even after personalizing the instructions, it does not seem the same. He is more technical, more generalized and honestly he feels emotionally distant,” wrote a member of the community in a row Entitled “Kill 4o is not innovative, it is cancellation”.
“Of course, 5 okay: if you hate the nuance and feel things”, another Reddit user he wrote.
Other Discussions They complained of slow responses, hallucinations and surprising errors.
Altman has promised to face these problems by doubling the GPT-5 rate limits for chatgpt plus users, improving the system that changes between models and allowing users to specify when they want to trigger a heavier and most capable “mode of thought”. “We will continue to work to get stable things and continue to listen to the feedback,” wrote the CEO on X. “As we said, we expected a little trick while we turn[ed] Out so many things at the same time. But it was a little more accidental than we hoped! “
Errors published on social media do not necessarily indicate that the new model is less capable than its predecessors. They can simply suggest that the brand new model has stumbled from several cases on board compared to previous versions. Openi refused to comment specifically the reason why GPT-5 sometimes seems to make simple mistakes.
The recourse has aroused a new debate on the psychological attachments that some users are formed with tram -trained chatbots to press their emotional buttons. Some Reddit users have rejected GPT-5 complaints as proof of unhealthy dependence on an artificial intelligence companion.
In March, Openi Research Published Explore the emotional ties that users are formed with its models. Shortly thereafter, the company issued a GPT-4o update after it has become too stapled.
“It seems that GPT-5 is less sicophantic, more” business “and less struggles”, says Pattie Maest, professor to Mit who worked on the study. “Personally I consider it a good thing because it is also what has led to disappointments, reinforcements of prejudice, etc. but unfortunately to many users as a model that tells them that they are intelligent and surprising, and that confirms their opinions and beliefs, even if even if [they are] wrong.”
Altman indicated in Another post on X That this is something with which the company has fought in the construction of GPT-5.
“Many people effectively use chatgpt as a sort of therapist or life coach, even if they wouldn’t describe it in this way,” wrote Altman. He added that some users could use chatgpt in ways that help improve their lives, while others may be “unwittingly popped up by their long -term well -being”.
Be First to Comment