The Bitter Lesson

Rich Sutton's 2019 essay arguing that the history of AI research consistently shows that general methods leveraging massive computation outperform methods that incorporate human domain knowledge — the 'bitter lesson' being that AI researchers keep re-learning this.

Articulated the empirical observation that scaling compute always wins over clever human-designed approaches — providing the intellectual framework for the scaling-first philosophy that drove GPT-4, AlphaFold, and the transformer era.