Abstract

This study analyzes the system, which is a key element of generative art, and discusses the AI image generative model in the genealogy and theory of generative art, in order to reveal the identity of generative AI-based art. Philip Galanter, a generative artist and theorist, defines generative art as any art practice where the artist uses a system, which is set into motion with some degree of autonomy, such as a computer program or procedural invention. The core of generative art is the use of systems, and the aesthetic goal of generative art that uses computers and programs is to create a “complex” system such as a genetic system or artificial life. The aesthetic value that generative art seeks to reveal lies in creating unexpected uniqueness while simultaneously using elements that increase from the highly ordered to the highly disordered. It is related to how extensively the system can accommodate mutations and noise, and secure diversity and complexity. Interestingly, the Diffusion Model(DM), which has recently appeared among generative artificial intelligence, is a case of the aesthetic use of complex systems, which generative art aims for. The diffusion model, which takes ideas from thermodynamics, can create a variety of unexpected images through forward diffusion, which intentionally injects noise, and a reverse restoration process, which removes noise through learning. As a result, the diffusion model, which exceeds the artist’s intuition while also bringing it closer to the artist’s desired result, can be an attractive tool for generative art.