One-Sentence Definition
Machine learning is a subset of artificial intelligence in which systems learn patterns from data and improve their performance over time without being explicitly programmed for each task.
How It Works
Traditional software follows rules a programmer writes. Machine learning flips this: you give the system data and the correct answers (labels), and the algorithm figures out the rules on its own. A classic example is email spam detection. Instead of writing hundreds of if-then rules, you feed the model millions of emails labeled "spam" or "not spam." The model identifies patterns -- certain phrases, sender behaviors, link structures -- and uses them to classify future messages.
There are three main paradigms. Supervised learning trains on labeled data (input-output pairs). Unsupervised learning finds hidden structure in unlabeled data, such as clustering customers by purchasing behavior. Reinforcement learning trains an agent through trial and error, rewarding desirable actions -- this is how DeepMind's AlphaGo learned to beat world-champion Go players.
Under the hood, most ML models are mathematical functions with adjustable parameters. During training, an optimization algorithm (often some variant of gradient descent) tweaks these parameters to minimize the difference between the model's predictions and the correct answers. The better the data and the more parameters the model has, the more nuanced patterns it can capture -- which is why modern large language models have hundreds of billions of parameters trained on terabytes of text.
Why It Matters
Machine learning powers the recommendations you see on Netflix, the fraud alerts from your bank, the voice recognition on your phone, and the drug-discovery pipelines at pharmaceutical companies. It is the engine behind virtually every AI product shipping in 2026.
The field is also accelerating. Techniques that required a dedicated research team five years ago -- fine-tuning a language model, training a custom image classifier -- can now be done by a single developer using open-source libraries like PyTorch, Hugging Face Transformers, and scikit-learn.
Key Takeaway
Machine learning is the practice of teaching computers to learn from data rather than follow hand-coded rules, and it is the technical foundation for nearly every AI application you use today.
Part of the AI Weekly Glossary.