Occam’s razor is about using the past to predict the future

Stuart Armstrong of LessWrong examines why Occam's Razor matters for prediction. The principle isn't just about favoring simplicity—it's about avoiding theories that fit past data perfectly but fail on future observations. Armstrong illustrates this with sun angles at different latitudes: a theory that merely lists every past measurement fits flawlessly yet predicts nothing new. The real insight is that simplicity prevents 'information leakage'—hidden assumptions and arbitrary choices that work on known data but won't generalize. That's why machine learning uses regularization, and why the razor remains central to prediction across all fields.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton