Recently, some users have complained about the use ofChatGPTmaybeGPT-4 A range of problems were encountered with the API, including slow responses, perfunctory answers, refusal to answer, and interruption of the session. In response, theOpenAIThe company acknowledged this fact and said it would fix the problems thoroughly and properly.
OpenAIexplained that the reason for these problems with GPT-4 was that the model had not been updated since November 11th. With the model becoming unpredictable, the company is working on fixing the problems.
OpenAIstated that they will update the GPT-4 model as soon as possible after a thorough offline evaluation and online A/B testing. Meanwhile, they also encourage users to actively feedback on the model performance. Behind the "laziness" of GPT-4 is the problem of model size and computational resources. As the model size continues to expand, the computational resources required for training and inference of GPT-4 also continue to increase, and there may be a lack of computational resources, resulting in the model showing a lazy attitude in certain tasks. There is also the problem of training data bias, which may exist in the training data of GPT-4, causing the model to fail to properly understand the requirements of the task in certain tasks. It also exposes flaws in the model structure, which may exist in GPT-4, resulting in its inability to reason and generate effectively in certain tasks.
This article comes from users or anonymous contributions, does not represent the position of Mass Intelligence; all content (including images, videos, etc.) in this article are copyrighted by the original author. Please refer to this site for the relevant issues involvedstatement denying or limiting responsibilityPlease contact the operator of this website for any infringement of rights (Contact Us) We will handle this as stated. Link to this article: https://dzzn.com/en/2023/2027.html
