Latent Autoregressive Composition Engine (LACE)

This is research project of my master’s thesis which explores the integration of AI-generated art into the artworld through the lens of George Dickie’s institutional theory, which posits that art is defined by its recognition as a candidate for appreciation within the artworld’s institutional context.

While generative AI has democratized artistic creation, AI-generated works often face skepticism regarding their artistic validity. To address this issue, we propose an artist-in-the-loop approach to generative models, such as Stable Diffusion, which aims to enhance the artist’s control and understanding of the model’s behavior.

By developing a tool that utilizes a human-computer interaction (HCI) approach to make the model more explainable, we seek to empower artists in their creative collaboration with AI. This approach not only refines the artistic capabilities of generative AI but also establishes a pathway for the ”artifactualization” of AI-generated art within the artworld’s institutional framework, ultimately bridging the divide between AI and human artistic expression in the digital age.

---

Download the full research thesis in PDF

Github RP: https://github.com/iamkaikai/LACE