In a hypothetical scenario, let’s say we are trying to use OpenAI’s model to generate dialogue that closely resembles that of Obi-Wan from Star Wars. Which approach should we focus on more - embeddings or fine-tuning? Alternatively, is it possible that neither is necessary and we could explore feasibility using the LangChain and few shot prompt instead?
I understand this is a challenging task, but I’m eager to discuss with anyone who is undertaking or considering similar initiatives.