Three ways to learn: with labels, without them, and from the data itself
Different methods use different kinds of feedback.
What is it?
In supervised learning a model learns from labelled examples (“this is a cat”). In unsupervised learning it looks for groups and patterns without labels. In self-supervised learning it creates its own tasks from the data, such as predicting a missing word.
An everyday example
Email that users mark as spam is supervised learning. An online shop that groups customers with similar purchases on its own is using unsupervised learning.
Why does it matter?
The kind of learning tells you how much human work was needed and where errors can creep in, for example in labels people assigned.
A common misconception
That these labels say how good a product is. They describe how it learned, not the quality of the result.
In more detail
What is supervised learning?
The model gets examples together with the correct labels. It learns the link between the input and the desired output.
What is unsupervised learning?
The system looks for patterns or groups in data without correct labels given in advance.
What is self-supervised learning?
The learning task is built from the data itself. For example, we hide part of a text and train the model to predict it.
Are there other kinds?
Yes. Semi-supervised, active, online, transfer and other forms of learning can be combined.
Remember: These differences describe how a model learns, not how good the product is.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.