In the past week, OpenAI has faced complaints from some users. These users noted that when using ChatGPT or GPT-4 APIs, they experienced a series of problems, such as slow responses, insufficient responses, refusal to answer, and interrupted sessions. OpenAI acknowledges this and promises to address these issues thoroughly and appropriately.
The cause of these problems, according to OpenAI, is due to the fact that the GPT-4 model has not been updated since November 11, causing the model behavior to become unavoidable. In particular, some users have reported that the GPT-4 model is "lazy" when performing tasks, such as providing only limited output when composing or processing ** requests, and telling users to complete the rest on their own.
It is understood that OpenAI admits that due to the long-term failure to update the model, the effectiveness of the GPT model on some prompt words has declined. However, the company emphasizes that the GPT-4 model still maintains good operational performance at the overall level. To address the current issue, OpenAI said it will update the GPT-4 model as soon as possible after a thorough offline evaluation and online A B test, and encourages users to provide feedback on the model's performance.
In addition, the stagnation of the update of the GPT-4 model is closely related to personnel changes within OpenAI. It is reported that the resignation of CEO Sam Altman and the superposition of holidays in Europe and the United States have led to chaos in the company's daily affairs. As a result, in addition to the regular update of GPT-4 being postponed, the GPT Store, which was originally scheduled to go live in early December this year for developers to share models, has also been postponed to early 2024.