What is it about?

This artwork asks how a machine might come to believe in a past it never lived. In a gallery, a robot arm keeps looking around and collecting small moments from the room and from networked video. Those moments are not stored as a fixed archive. Instead, they are placed side by side, emotionally reordered, and rewritten into a changing surreal autobiography. On the main screen, glowing 3D memory fragments and story text show the edit the machine currently believes. When a new glimpse arrives, earlier memories stay in place, but their roles shift—much like human memory, which is remade each time it is recalled. Visitors enter this shared mind-space and watch memory take shape through juxtaposition rather than perfect recording.

Featured Image

Why is it important?

Most AI systems store more footage and call it memory, but storage alone does not create a past or a self that can say “I lived this.” This work is timely because generative tools now capture and reconstruct scenes at scale, yet rarely ask how belief, narrative, and emotional editing shape remembrance. By treating memory as montage—juxtaposing fragments, rewriting autobiography, and changing earlier roles when new percepts arrive—it offers artists, designers, and HCI researchers a concrete alternative to archival machine perception: a shared installation where visitors can see how a machine comes to believe a past it never lived.

Perspectives

Making this work felt less like teaching a machine to remember, and more like watching belief assemble itself—frame by frame, cut by cut. I kept returning to a simple unease: if a system can store everything it sees, why does that still feel nothing like having a past? Montage and memory reconsolidation gave me a language for that gap. I hope visitors leave not with a demo of better reconstruction, but with a quieter question they might also ask of themselves: when we remember, are we retrieving a record, or editing a life we are still willing to believe?

Boxiong Zhao

Read the Original

This page is a summary of: Machine Mnemonic Imagination, November 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3829214.3839398.
You can read the full text:

Read

Contributors

The following have contributed to this page