SoTA

KAIST XD Lab (Experience Design Lab), KAIST (Korea Advanced Institute of Science & Technology)

Artwork Description:

State-of-the-Art is an academic term describing AI models with exceptional performance, so exceptional that they are considered works of art. SoTA investigates this metaphor further, drawing a parallel between AI’s evolution and the history of Western Art. Early AI architectures (McCulloch-Pitts, CNNs, etc.) were inspired by biological neural structures, mirroring pre-modernist representational art. Modern Transformers (LLMs, GPTs, etc.), however, favour self-referential computation over human inspirations and biological metaphors. Just as photography pushed art toward Greenbergian Modernism’s “art for art’s sake,” AI is evolving toward “AI for AI’s sake,” a vast computational sublime beyond human comprehension. SoTA unfolds in two stages: the representational Frontend and the abstract Backend. As the audience scans the QR, they access the custom-developed mobile website and scroll through the list of 118 AI models. In real-time, surrounding projections and screens within the installation render each model’s 3D architecture, visualising the neural network. Tilting the phone rotates the visualisation synchronously, creating an embodied sensation of residing inside the AI. The experience shifts unexpectedly upon disengagement. When the audience locks the phone or exits the browser, all screens change abruptly. Colourful, representational 3D visualisations vanish, replaced by a monochromatic flux of lines. The user’s name floats amid real-time GPT-generated text, connected by curved lines calculating cosine-similarity between LLM token embedding vectors, resembling Jackson Pollock’s action paintings. This Frontend-Backend dual system, accompanied by the abrupt transition inspired by a Brechtian Verfremdungseffekt, questions: Where is the state-of-the-art of our age heading? Do humans still have control over this abstraction?

Project Collaborators:

Artwork KAIST XD Lab (Experience Design Lab) Lead Artist: Jeanyoon Choi Software Engineer & Visualisation: Jeanyoon Choi Sound Designer: Euan Kang 3D Designer: SeJoon Park UI Designer: Minhyeok Seo Advisor & Director: Prof. Yiyun Kang Research Lead Researcher: Jeanyoon Choi User Study & Findings: Intae Hwang Research Assistant: SeJoon Park Research Direction: Prof. Hyungjun Cho Writing Revision: Heejae Bae Corresponding Author: Prof. Yiyun Kang Received ACM DIS 2026 Honourable Mention Award. Funding & Venue KAIST Art Museum