Hand-computable AI? Markov chains plus polynomial feature fits for POS tagging

cephaloform · x · 2026-09-16

Asked whether AI could technically be computed by hand since LLMs are just math, cephaloform describes a fun toy approach: ensembling Markov chains for part-of-speech tagging with linear equation fits over polynomial combinations of a random 1D projection of semantic features, using 4 words of context. Trains in about an hour and is fun to sample.

Related event: Markov Chains Plus One-Dimensional Projections Enable Hand-Computable AI(3 posts)→

Original post →

More from Fun

Fun channel →