
Razib Khan's Unsupervised Learning Monologue: Race - genetics, history and sociology
17 snips
Apr 23, 2026 A deep monologue exploring the tangled meanings of race across genetics, history and sociology. It traces ancient observations, African origins, early population splits and out-of-Africa dispersals. It contrasts typological thinking with population genetics and explains how modern DNA markers and ancestry tests can classify geographic and population structure.
AI Snips
Chapters
Transcript
Episode notes
Race Is A Constructed Classifier
- Race is a socially constructed classification rather than a fundamental natural kind.
- Razib Khan compares race to a table or species: useful human-made labels that organize real variation but are not elementary particles.
Lewontin's Result Doesn't Preclude Classification
- Lewontin's finding that most genetic variation is within groups was accurate for single loci but misleading for multi-locus classification.
- Razib Khan explains allezyme-era results (1970s) versus later multi-marker classifiers that can distinguish populations reliably.
A Few Hundred Markers Can Identify Ancestry
- With a few hundred highly variable markers, geneticists can classify continental and subcontinental ancestry with very high accuracy.
- Neil Risch and others showed early 2000s microsatellite/SNP work made classifiers routine for groups like African-American, Latino, and European.
