Comparative Prompting Across Models

This unit surveys major generative and encoder models (ChatGPT, DALL·E, CTRL, T2T, BERT, Tacotron, MuseNet), focusing on how prompts function across text, image, audio, and music modalities. Students will compare capabilities, limitations, ethical considerations, and future directions of each model, building on foundational prompt principles and preparing for applied prompt design and evaluation in subsequent units.

Learning Objectives

• analyze how prompts influence output behavior and control mechanisms across at least four models (e.g., ChatGPT, DALL·E, BERT, Tacotron) and across text, image, and audio modalities
• apply model-specific prompting techniques to produce targeted outputs in at least three modalities (text, image, audio/music) and document the prompt–output relationships
• evaluate the technical limitations, failure modes, and ethical or societal implications of each model and propose mitigation or responsible-use strategies
• demonstrate selection of the most appropriate model and prompt strategy for given real-world use cases, justifying choices based on capability, constraint, and ethical considerations