New book Vector Media dispels 'generative' AI, proposes 'neural exchange value'
round · x · 2026-09-30
Leonardo Impett and Fabian Offert's new book Vector Media (University of Minnesota Press, 176 pages, April 2026) argues for dropping the notion of "generative" AI.
- Neural networks dissolve all media into vector space — a universal space of commensurability — constituting a technical ideology of epistemic reduction.
- Under this regime, bias is not just about what is represented but about the logic of representation itself.
- Analyzing techniques behind DALL-E, Midjourney, Flux and Stable Diffusion, the authors propose "neural exchange value": cultural artifacts gain value not through meaning or context but through their capacity to function as vectors — what matters is not what something is, but that it can be embedded.
- The book draws on Phil Agre's critical technical practice to show how AI systems embed new epistemologies of media beneath their architectures.
More from AGI Musings
- Agent Economy Deep Dive: 30+ Public Companies Rated for AI Agent Disruption Risk — RihardJarc · 2026-09-30
- Paras Chopra: Post-work society seems inevitable as we outsource thinking to AI — paraschopra · 2026-09-30
- Only 2.6% of Agentic Tools Match Existing Human Work Tasks, Cohere Study Finds — Cohere_Labs · 2026-09-30
- Researcher debunks AI consciousness claims: inference is stateless, prompts aren't people — gerardsans · 2026-09-30
- Indie iOS dev: "It's over" — Claude Code wrote 49 of his 50 app intents, ending hand-coded work — JordanMorgan10 · 2026-09-30
- AI Can Write Code, Docs and Tests — But Not Judgment, Says Engineer — arpit_bhayani · 2026-09-30