WORK
Gramsci's Nightmare (Zuckerman Lecture)
Ethan Zuckerman's 2025 University of Copenhagen lecture identifying the specific mechanism through which large language models lock hegemonic values into computational infrastructure — compounding them in ways Gramsci did not anticipate.
Zuckerman's lecture provides the sharpest contemporary application of Gramscian analysis to AI infrastructure. The argument: LLMs are built by compressing a civilization's worth of culture into opaque matrices of linear algebra. The values embedded in those matrices — assumptions about what counts as knowledge, what counts as reasonable, what counts as neutral — are the values of the particular population that produced the training data: disproportionately English-speaking, disproportionately Western, disproportionately the product of the early twenty-first-century open internet. These values do not announce themselves as particular. They present themselves as the model's general intelligence. The neutrality is the ideology.
In The You On AI Field Guide
The most cited passage crystallizes the argument: "AI automates this reinforcement — the WEIRD values of the texts that build this new form of intelligence are not just common sense, they are how the machine knows how to answer questions and produce text… and as AI feeds on the texts it creates, an ouroboros