User Friendly and Adaptable Discriminative AI: Using the Lessons from the Success of LLMs and Image Generation Models