How AI Solves Problems
Rule-Based vs Learned Systems
6 min+55 XPStep 1/10
Two ways to build a deciding machine
Rule-based systems encode logic explicitly: experts write if temperature > 38°C and cough, suspect flu. The result is a flowchart — transparent, predictable, and limited to exactly the cases somebody anticipated.
Learned systems are never handed the logic. They get examples — labeled emails, recorded games, millions of sentences — plus a measure of success, and discover the patterns themselves. The 'rules' end up as millions of tuned numbers that nobody wrote.
analogy
Rule-based is a recipe card; learned is a chef who tasted ten thousand dishes. The card is precise but brittle. The chef handles dishes nobody wrote a recipe for — but can't always explain the pinch of salt.