Twenty-First International Conference on Technology, Knowledge, and Society

Abstract

Various concerns are associated with artificial intelligence (AI) tools, including irrelevant or off-topic responses, contextual misunderstanding, and lack of control. Based on an extensive literature review, this paper discusses the concept of prompt engineering for effective communication with large language models (LLMs), such as GPT-3 upon which ChatGPT was built. We demonstrate the benefits of effective prompt engineering processes in AI tools, among them improving the model performance, increasing the output quality, and handling bias issues. We also describe several key principles for the optimization and accuracy of the AI models’ answers to user queries. The results reveal that concise, clear, and progressive prompting is among the best practices for prompt engineers. In the developing era of intelligent interactions, future researchers would do well to focus on studying the future developments in the field of prompt engineering, as well as clarifying the modes of their influence on society.