Sam Altman, the CEO of OpenAI, the company that developed ChatGPT and its refined successor, GPT-4. Jason Redmond/AFP via Getty Images
Users of OpenAI’s new GPT 4 model have complained about performance degradation in recent weeks.
OpenAI admitted that GPT 4 performance can be worse on some tasks.
“There may be some tasks where performance degrades,” OpenAI explained in a blog post.
This is a machine translation of an article by our US colleagues at Insider. It was automatically translated and checked by a real editor. We welcome feedback at the end of the article.
OpenAI, the company that developed ChatGPT, said GPT-4’s performance might be worse for some tasks, as users have complained that the model is “lazy” and “dumber” compared to previous versions.
“While most metrics have improved, there may be some tasks where performance is degrading,” OpenAI explained in a blog post on Thursdaywhich announced new updates.
OpenAI’s GPT-4 users have been complaining about the inferior performance compared to the earlier reasoning capabilities over the past few weeks.
read too
You don’t have to train every day: 2.5 hours of sport per week is enough
Some vented their frustration on Twitter and in the OpenAI online developer forum Air. Among other things, they reported weaker logic, more erroneous responses, losing track of information provided, problems following instructions, forgetting parentheses in basic software code, and the fact that only the last prompt is saved.
“The current GPT-4 is disappointing,” wrote a developer using GPT-4 to code functionality for their website. “It’s like driving a Ferrari for a month and then it suddenly turns into an old, beat-up pickup truck. I’m not sure I want to pay for it.”
Read the original article in English here.