Faculty of Arts & Science

Digital Futures

Yixuan Xiao

Life Noise

This project proposes an anti-efficiency AI service that engages with the “noise” of everyday digital life. It works with discarded personal data—such as blurred or imperfect images—and generates abstract, poetic textual responses that evoke the texture of lived experience rather than enhancing or summarizing it. By refusing optimization, insight, or emotional diagnosis, the project challenges productivity-driven design logics and questions the assumption that personal data must be clarified, improved, or made meaningful. As a speculative and experiential system, it allows unstructured traces of life to exist without justification.Please click the web icon below to experience it.

“This project explores an alternative paradigm for AI-driven services by proposing a system that deliberately resists efficiency, optimization, and interpretation. Working with discarded personal data, the project shifts attention away from clarity and meaning toward ambiguity and presence. Rather than extracting insight or constructing narratives, the system generates sparse, contemplative textual outputs that evoke the texture of lived experience without resolving it.Conceptually, the work challenges dominant assumptions in digital culture that equate care with productivity, and data with something to be refined or improved. It reframes incompleteness and indeterminacy as valid states of living, allowing personal traces to exist without justification or evaluation. Through strategies such as delay, disappearance, and non-archival processing, the project constructs a non-performative interaction model that resists accumulation and feedback.Positioning AI as a passive mediator rather than an analytical agent, the work offers a quiet form of companionship—one that does not seek to understand, diagnose, or optimize, but instead makes space for what is often overlooked. In doing so, it invites users to reconsider their relationship with digital systems, and to encounter everyday life not as something to be processed, but simply as something that unfolds.This system is developed in Visual Studio Code and Vercel, integrating both a large language model and an image generation model within framework coded by multiple programming languages. Through computational capture of visual inputs—such as pixel composition, color distribution, texture, and light—the system sees images not as subjects to be identified, but as carriers of lived traces. These sensory dimensions are translated into abstract textual responses, enabling the AI to engage with the subtle “texture” of everyday life rather than extracting explicit meaning or categorical information.”