The Impact of Prompt Pattern Catalog on Prompt Engineering with ChatGPT

Abstract ―​

ChatGPT and other new conversational AI systems showcase impressive natural language capabilities. However, their potential is constrained without effective prompts encoding human goals and knowledge. Manual prompt engineering remains challenging and ad hoc. This article proposes shared prompt pattern catalogs to accumulate expertise and systematize prompt engineering. Catalogs would collect crowdsourced prompts for diverse applications along with guidance on usage. They would establish conventions and vocabularies for discussing prompts, enabling organization and discovery. Expert curation would ensure accuracy, safety, and structure. Related discussion platforms would provide context missing from raw prompt examples. Prompt pattern catalogs can enhance prompt engineering through collaboration, transparency, and standardization. They offer significant productivity gains over isolated efforts. But responsible governance is required as catalogs gain influence over AI systems. Investing in shared prompt resources will be key to safely unlocking the power of AI. This article explores the envisioned benefits of catalogs, while considering risks and implementation challenges. It offers a roadmap for steadily advancing the science of prompt engineering.