AI Is Not Lying Because of Entropy

Wed, 01 Jan 2025 00:00:00 +0000

A lot of contemporary criticism of AI leans on “entropy” as if it were a physical proof that statistical systems must inevitably decay into falsehood. But that reading of entropy is not actually in thermodynamics, information theory, or machine learning. It is a cultural inheritance that predates all three.

The second law of thermodynamics is a narrowly scoped claim about energy distributions under specific constraints. It does not say that systems become worse, less meaningful, or less truthful. The move that turns entropy into decay, degeneration, or moral failure enters through nineteenth-century rhetoric, especially in Victorian Britain. When figures like William Thomson (Lord Kelvin) popularized the “heat death of the universe,” the mathematics concerned gradients and work; the language framed this as the universe “running down” toward futility. That framing drew directly on Christian eschatology: an ordered creation falling toward an exhausted end state. Entropy became a story about destiny, not just a state function.

CulturalCriticism on emsenn.net

AI Is Not Lying Because of Entropy