Software / Machine Learning

Reading at Scale

A language model predicts likely text, and at scale that becomes summarizing, translating, drafting, and writing code. It is a strong reader and an unreliable witness: useful for synthesis across more material than one person can hold, and trustworthy only once its claims are checked against a source or a test. Spw's theory grew partly against model-written syntheses of new STEM work, which is why this site pairs every model-assisted change with a check.