Generative AI and ChatGPT

ChatGPT kept telling me that it is just a language model and is not designed to remember even basic stuff such as information about types of pizza crust or toppings. However, in other interactions, I have found that ChatGPT is clearly knowledgeable about pizza ordering. So, it seems that prompt engineering is key to getting desired responses from ChatGPT.

ChatGPT also told me that it is not capable of zero-shot learning but can be fine-tuned. However, the underlying GPT-3.5 model is capable of zero-shot learning. Further, you would expect a large-scale general-purpose chatbot like ChatGPT to recall basic info about common things like pizza ordering and be able to have an intelligent conversation pertaining to the same.

I assume that OpenAI is going to release a commercial API for ChatGPT. However, if it is not capable of zero-shot learning, I am not sure how it can be effectively used as the basis for a task-specific conversational commerce chatbot. Perhaps, that is the not the intended domain of application. Other than prompt learning, how can you fine-tune the ChatGPT API (when released) for applications such as conversational commerce?