CS224n week 1 maps NLP's four eras, from rules and stats to scaling and Shannon

stanfordnlp · x · 2026-10-12

A CS224n student summarizes week one: NLP has gone through four eras — rules lost to statistics, statistics lost to scaling, vindicating Sutton's Bitter Lesson.

Re-reading Shannon's 1948 paper after class, the noisy channel model still underlies machine translation; the post's images include the 1954 Georgetown-IBM machine translation demo coverage and Shannon's noisy channel diagram. It also notes Anthropic named Claude after Shannon.

Original post →

More from Companies & People

Companies & People channel →