Unreliable Past: Constructing Memory with Generative AI

Abstract The author examines generative AI as a new type of “memory machine,” exploring its role in constructing and visualizing both personal and collective memories. Through theoretical analysis and artistic practice, he investigates how AI-generated images combine historical documentation with aesthetic idealization, creating representations that are simultaneously familiar and unreliable. The author contextualizes this phenomenon within the history of media technologies and demonstrates its artistic potential through his recent series Ideal City, which uses AI to reimagine urban environments from his childhood. The work reveals how AI’s technical limitations can become meaningful metaphors for memory’s inherent unreliability.