A large technology company has already invested billions of dollars to pre-train a massive, general-purpose language model. A new product team wants to use this model to create a specialized chatbot for medical professionals. Analyze why adapting the existing model using a curated dataset of medical questions and answers is a more cost-effective strategy for this specific goal than developing and training an entirely new model from the ground up. In your analysis, break down the key differences in resource requirements between the two approaches.