On Tuesday, Openai CEO Sam Altman said that the company is “rollback” to the latest update for the default AI model Powering CHATGPT and GPT-4O after dissatisfaction with strange behavior, especially severe Sycophancy.
Altman wrote in the post on X, “(W) began to roll back the latest update for GPT-4O last night.” (I) TE received 100% roll back for free CHATGPT, and it will be updated again when it is completed for a paid user today.
During the weekend, social media users blamed the updated models that arrived at the end of last week, allowing Chatgpt to be overly verified and agreed. It became a meme quickly. The user posted a screenshot of CHATGPT while applauding all kinds of problems, dangerous decisions and ideas.
Altman acknowledged this issue on Sunday and said that Openai will modify “ASAP” and “ITS learning” at some point.
