Machine learning (ML) is a branch of artificial intelligence in which computers learn patterns from data and improve at a task through experience, rather than following rules written by a programmer. You give an ML system examples. It finds the patterns. Then it uses those patterns to make predictions or decisions about new data it has never seen.
Key takeaways
- Machine learning is the main method used to build modern AI.
- The three classic types are supervised, unsupervised and reinforcement learning.
- The core tasks are classification, regression and clustering.
- The goal is to generalise: perform well on new data, not just the training data.
Machine learning explained in simple terms
Traditional software works like a recipe. A programmer writes every step: “if the email contains these words, mark it as spam”. Machine learning flips this. You show the computer thousands of emails already marked spam or not spam, and it works out the rules itself. When spammers change tactics, you retrain it on new examples instead of rewriting the code.
Computer scientist Tom Mitchell gave a widely used definition in 1997: a program learns from experience E, with respect to a task T and a performance measure P, if its performance on T, as measured by P, improves with experience E.
Where the term came from
Arthur Samuel of IBM popularised the term “machine learning” around 1959 through his work on a checkers program that improved by playing games against itself. The field grew slowly for decades, then accelerated in the 2010s as data, computing power and neural network methods improved together.
How does machine learning work?
- Collect data relevant to the problem.
- Prepare it: clean errors, handle missing values and apply feature engineering.
- Split it into training, validation and test sets.
- Train a model: an algorithm adjusts the model to reduce a loss function, often using gradient descent.
- Tune and evaluate, watching for overfitting and underfitting.
- Deploy and monitor through an ML pipeline, retraining when data changes.
Types of machine learning
| Type | Learns from | Typical use |
|---|---|---|
| Supervised learning | Labelled examples with correct answers | Spam detection, price prediction |
| Unsupervised learning | Unlabelled data | Customer segmentation, anomaly detection |
| Semi-supervised learning | A few labels plus many unlabelled examples | Document tagging with limited labels |
| Self-supervised learning | Labels created from the data itself | Training large language models |
| Reinforcement learning | Rewards from trial and error | Games, robotics, recommendation tuning |
Machine learning vs AI vs deep learning
Artificial intelligence is the broad goal of machines doing intelligent tasks. Machine learning is one way to reach that goal, by learning from data. Deep learning is a type of machine learning that uses neural networks with many layers. Think of three nested circles: deep learning sits inside machine learning, which sits inside AI.
Where machine learning is used
ML runs quietly behind search rankings, product recommendations, fraud checks, credit scoring, demand forecasts, medical imaging tools, translation, voice assistants and chatbots. In business, the most valuable uses are often unglamorous: predicting churn, spotting defects or forecasting stock.
Limitations
Machine learning needs enough good data. It can learn bias from historical records. Complex models can be hard to explain. And models degrade when the real world changes, a problem called data drift. Machine learning is a powerful tool for finding patterns, but it does not replace clear goals, domain knowledge and human judgement.