The First Rule of Machine Learning: Start without Machine Learning

Eugene Yan · 5 years ago · 49

Eugene Yan argues that teams should solve problems using simple heuristics and rules before applying machine learning, citing similar advice from Google's ML rules and interviews with practitioners at major tech companies. Simple solutions like regex rules, exclusion lists, and manual heuristics often match or outperform complex ML models while requiring less effort and maintenance. ML should only be adopted once non-ML baselines become too complex to maintain and robust data pipelines with high-quality labels are established.