This release significantly increases the AI model's accuracy, possibly by 1,000 times, by allowing you to use a completely different technique for creating your own chatbot, based upon a combination of "embeddings" and "text-davinci-003", where we use embedding to generate a "context" which we then supply to OpenAI to use as its foundation to answer your questions.
It basically combined semantic search through your snippets collection, with ChatGPT3, generating a response to your answer to for all practical concerns is at least 1,000 times more accurate than what you get if you fine-tune your own model.