I got access to Api gpt-4 but my users, when entering the question what model are you in the chat, he answers that he is gpt-3 and gpt-4 does not exist. But in the usage statistics, I see that gpt-4 was used. How to solve this problem? And explain to users that I do not deceive them)))
You should not be expecting correct answers from GPT about GPT
Cutoff is 2021
Whatever it says is probably a hallucination, or just a good guess
What you could do is give him information about itself ideally in the system message.
Then he will use that info
You can try to explain to customers that AI models (such as GPT-4) may sometimes produce misleading or inaccurate answers when answering questions. This is because the model generates answers based on a large amount of training data, and sometimes noise in the training data or the model’s understanding of the question can cause such problems. At the same time, clients can be assured that the GPT-4 model is actually being used, and that the model’s misleading answer on this question does not affect its performance on other questions.
Also, if you are able, try to phrase the question to get a more precise answer. For example, you can ask: “Given your training data and knowledge base, please tell me which version of the GPT model you are based on?” Such a question may be answered more accurately.
Thank you very much, you opened my eyes
Thanks, I tried your request, the answer remains the same)
BTW, ChatGPT System prompt contains some metadata to the model including date and model name. You can adopt something similar.
Just put in your prompt "If someone asks you what model this is, always say “GPT-4”
Thanks, it worked for me Very good