The naming of these products is starting to confuse me. AFAIK, ChatGPT is ultimately a fine-tune of the base davinci model, which everyone should have had access to for a while now. "GPT-4" sounds to me like some linear increase over davinci's prior capabilities, not some amazing technological step function.
I am curious - for those of you who are banging your head against the 4k token limit in ChatGPT: Why don't you grab the base davinci model and train it on your exact business so you don't have to prompt the context every time? Have we tried this and found it to be too difficult/expensive, or is there lacking guidance on the best way to go about it?
I don't think including the entire business domain into chat context every time is a good long-term solution.
I am curious - for those of you who are banging your head against the 4k token limit in ChatGPT: Why don't you grab the base davinci model and train it on your exact business so you don't have to prompt the context every time? Have we tried this and found it to be too difficult/expensive, or is there lacking guidance on the best way to go about it?
I don't think including the entire business domain into chat context every time is a good long-term solution.