Vector Media
Neural networks are designed to dissolve all media into the vector space—a universal space of commensurability. In Vector Media, Leonardo Impett and Fabian Offert parse theories of automatic vision to trace contemporary artificial intelligence’s technical ideology of epistemic reduction, where sensory data is turned into abstracted forms of meaning. Under this regime, bias is not just a question of what is represented but of the logic of representation itself. Drawing on Phil Agre’s notion of a critical technical practice, Vector Media reveals how artificial intelligence systems embed new epistemologies of media beneath the surface of their architectures.
Analyzing the techniques underpinning large multimodal artificial intelligence models like DALL-E, Midjourney, Flux, or Stable Diffusion, Impett and Offert offer the concept of neural exchange value: the value cultural artifacts acquire not through meaning or context but through their capacity to function as vectors. In such a system, commensurability becomes a condition of existence: what matters is not what something is but that it can be embedded. Rather than focusing solely on datasets, Vector Media proposes a critical study of vector spaces—and the machine cultures they produce—as a necessary complement to prevailing approaches in AI critique.
