CONCEPT
The Knowledge Problem in AI
Hayek's knowledge problem—that critical information is dispersed and contextual—applied to LLMs that aggregate expertise while stripping the situated context that made it valuable.
The knowledge problem in AI is the extension of Friedrich Hayek's 1945 framework into the age of large language models. Hayek argued that the knowledge required to coordinate a complex economy is dispersed across millions of individuals, each possessing fragments of contextual understanding that cannot be centralized without losing value. AI systems aggregate human expertise at unprecedented scale—training on the outputs of millions of programmers, writers, lawyers, doctors—extracting patterns from dispersed knowledge and making those patterns available through a single interface. The Hayekian question is whether this aggregation preserves knowledge or transforms it into something categorically different: patterns that match the form of expertise without possessing its contextual substance. The pattern "antibiotics cure infections" is useful; the situated judgment "this patient needs this antibiotic at this dose given this history" is knowledge. AI excels at the first, struggles with the second.
In The You On AI Field Guide
Hayek distinguished between scientific knowledge (general principles) and knowledge of particular circumstances (local, timely, contextual). The