In Atlas of Perception, Tom White translates the hidden structures of machine vision into expansive visual fields. Trained as both an AI researcher and an artist, White draws directly from the latent space of neural networks, revealing how artificial systems perceive not objects, but relations—patterns that unfold radially rather than linearly. Circuit (Latent 3411) and Root (Latent 0006) visualize two such perceptual primitives: one suggestive of engineered networks, the other of organic growth. At first glance, each resembles its real-world referent, yet closer inspection reveals something unstable and synthetic.
These works function less as discrete images than as continuous systems, with lines that connect, proliferate, and loop. Presented as circular, unbounded compositions, they mirror the non-hierarchical structure of neural networks themselves. White’s project asks what it means to see through a machine, proposing vision as a grammar of recurring elements rather than a catalog of things. It invites viewers to inhabit perception as a dynamic, relational field.