Instead of writing rules for every situation, you show a computer examples and let it find patterns. That is machine learning — algorithms that learn from data to make predictions. A spam filter learns from millions of labeled emails. A recommendation engine learns from what you watched and what you skipped.

Machine learning has three main paradigms. Supervised learning uses labeled data. Unsupervised learning finds structure in unlabeled data. Reinforcement learning trains agents through rewards and penalties. Deep learning uses neural networks with many layers to handle images, speech, and text.

Common applications

• Image recognition: faces, objects, medical scans
• Natural language: translation, chatbots, sentiment analysis
• Recommendation: products, movies, music
• Fraud detection: credit cards, insurance claims
• Autonomous systems: cars, drones, robots

Machine learning is powerful but imperfect. Models can be biased, brittle, and hard to interpret. They need large amounts of clean data. They fail in ways humans find obvious. The field is advancing fast. The guardrails are not.