Journal for Geometry and Graphics 28 (2024), No. 2, 213--227
Copyright by the authors licensed under CC BY SA 4.0
Analysis of AI Prompts for Introductory Education in Generative Art
Kunio Kondo
University of Technology, Hachioji, Tokyo, Japan
kondo@stf.teu.ac.jp
Hiroyuki Nagayoshi
Design Unviersity, Kobe, Japan
nagayoshi-h@kobe-du.ac.jp
Makoto Nagata
Design Unviersity, Kobe, Japan
nagata-m@kobe-du.ac.jp
Junichi Kanebako
Design Unviersity, Kobe, Japan
kanebako-j@kobe-du.ac.jp
Generative Art is creating CG images by designing and implementing image generation algorithms as programs. This method requires a solid understanding of algorithms and programming, which can be challenging for students and designers who do not have these skills. Therefore, our research proposes guidelines for writing prompts that support precise descriptions of image generation algorithms. The research involves three exercises: (1) Describing algorithms in natural language and generating and evaluating the resulting images. (2) Analyzing artworks created through Generative Art methods to derive algorithm descriptions, 3. Generating CG images using prompts. These exercises allow students to describe procedural steps in natural language, enabling AI to generate p5.js programs without programming expertise. Through analysis of the images produced and student feedback, the study identified key principles for designing effective prompts, essential considerations for program development, and guidelines for accurate CG image generation. While AI’s interpretive capabilities improve, students’ logical thinking and precise prompt-writing skills remain crucial for effective human-AI collaboration.
Keywords: Generative art, introductory education, ChatGPT3.5, AI prompts, algorithms.
MSC: 68T01; 97G99.
[ Fulltext-pdf (2266 KB)]