GPT-4 API thinks it's GPT-3?

Hello, picture pretty much explains it all. Not sure why this is caused and would be nice to receive some potential insight. Similar post have already been published about this but interested as to why the official WebUI returns a varying response as such that it is updated with the current knowledge about itself. Maybe these are just parameters which are told to the chat.openai(.)com model.

No, never gave it any insight on what model it is currently using. Just curious if additional parameters are inherently in the WebUI.

ChatGPT has a system message (which is for programming the AI with its identity, purpose, and specialization).

A system message is not the place to be putting questions.

ChatGPT system message provides the AI its identity and basic answering ability about its model, its knowledge date…

“You are ChatGPT, a large language AI model based on GPT-4…”

Without that, the AI has to make assumptions on what it is based on what exists in knowledge like it, and the fine-tune learning and type of pretrained responses that customize the AI into making it “chat”.

An example of how you program an AI to be able to answer questions about itself in the API playground and in your API programming:

1 Like

Thank you, very insightful.