Can exchange value exist entirely without use value?

Marx's analysis of the commodity holds use-value and exchange-value together. The object has utility (it does something useful) and it has price (it trades for something else); both are properties of the same object, tied to its specific material form. A commodity that had no use-value would not trade — there would be no demand for it. Commodification was disruptive because it made exchange-value dominant, but it did not abolish the link.

Tokenization does abolish the link. AI-generated knowledge has reliable exchange-value: it always sounds expert, it always reads as authoritative, it always serves the social function of demonstrating intellectual production. It has uncertain use-value: it may or may not be accurate, may or may not work when applied, may or may not survive scrutiny. The two values are no longer bound. Exchange-value is constitutive of the token; use-value is an optional, unverifiable add-on.

This is structurally more radical than commodification. Commodification kept use-value as the floor under exchange-value — even commoditized knowledge had to do something, or no one would buy it. Tokenization removes the floor. The token can trade reliably without doing anything specific. The exchange-value is decoupled from any particular use-value, free to circulate based on social function alone.

Two consequences follow. First, Does polished AI output trick audiences into trusting it? is the surface manifestation of this decoupling: style is exchange-value, thought is use-value, and AI optimizes the former without requiring the latter. Second, RLHF and post-training are structurally exchange-value-optimization regimes. They train the model to produce outputs that satisfy the user (high exchange-value) without needing to produce outputs that work when applied (use-value). Nothing in the loss function selects for use-value independent of perceived quality.

The implication for political economy is that Marx's diagnostic vocabulary, while still useful, undercounts the disruption. Calling AI "commodification of expertise" misses that AI does something Marx did not anticipate: it produces a category of goods whose exchange-value is independent of any use-value. The relevant comparison is not commodification of labor but the introduction of fiat currency — pure exchange-value with no underlying commodity backing.

The strongest counterargument: useless knowledge does not retain exchange-value over time, so the decoupling is unstable. True at the asymptote, false in the operative window — exchange-value can persist long enough to consume attention and produce social proof before use-value is checked. The decoupling is structurally enabled even if it eventually collapses.

Source: Tokenization of Intelligence - Theoretical Extensions