How to initialize GPT-4 with a custom instruction only once and use it after that

[tropic stream]

currently, I use gpt4 api in the following manner: I need to send the custom_instruction every time which is long and used lots of tokens. Is there a way to initialize gpt4 once with the custom instruction and keep using it?

    response = openai.ChatCompletion.create(
        model="gpt-4",
        messages = message,
        temperature=0,
        max_tokens=1000,
        frequency_penalty=0.0
    )```

[keen canopy]

I don't believe you can, at least perhaps not until fine tuning for GPT-4 becomes available then you might be able to implicitly implement the system prompt as part of the fine tuned model