What is machine learning?
Machine learning (ML) is a way of building software that learns patterns from examples instead of being given every rule by hand. In traditional programming, a developer writes the exact rules: "if the email contains these words, mark it as spam." In machine learning, you instead show the system thousands of emails already labelled spam or not spam, and it works out the patterns itself. The result is a model, a trained piece of software that makes predictions about new data it has never seen.
In plain words
Think of machine learning like teaching a child to recognise dogs. You do not hand them a checklist of fur length and ear shape. You point at dogs, say "dog" a few hundred times, and eventually they recognise one they have never met. Machine learning works the same way: show enough examples, and the system learns to recognise the pattern, even on new cases.
How it works in one paragraph
ML has two phases. First comes training: you feed the system lots of example data, and it adjusts itself until its predictions match the known answers. Then comes inference: you give the trained model new, unseen data and it makes a prediction. The more relevant and clean the training data, the better the predictions. Crucially, the model never stores rules a person wrote. It stores patterns it found in the data.
Where you already meet it
- Spam and fraud filters that learn what suspicious messages and transactions look like.
- Recommendations on Netflix, Spotify, and e-shops, based on what similar people chose.
- Forecasting demand, churn, or stock levels from past behaviour.
- Modern AI tools. Large language models behind ChatGPT and Claude are machine learning at large scale.
For a business, ML matters when you have lots of historical data and a decision that repeats often. It turns that data into predictions you can act on, rather than rules someone has to maintain by hand.
Common pitfalls
- Garbage in, garbage out. A model learns whatever is in the data, including its mistakes. Biased or messy data produces biased or unreliable predictions.
- It predicts, it does not understand. ML finds correlations, not causes. A model can be confidently wrong, especially on cases unlike anything it was trained on.
- Drift over time. The world changes, and a model trained on last year's behaviour slowly gets worse. Models need monitoring and retraining.
- Not every problem needs it. If a simple rule solves it, a rule is cheaper, clearer, and easier to trust than a model.
Related articles:
- What is AI? - The bigger field that machine learning sits inside.
- Machine learning vs. deep learning - How a powerful subset of ML differs from the rest.
- What is a neural network? - The structure behind today's most capable models.
Want to stay one step ahead?
Don't miss our best insights. No spam, just practical analyses, invitations to exclusive events, and podcast summaries delivered straight to your inbox.