How machines learn
Unsupervised learning is machine learning on data that has no labels, where the model must find structure, groups, or patterns on its own.
Sometimes you have piles of data and no answers. No one has tagged which customers are similar or which songs belong together. Unsupervised learning hands the model the raw data and asks: what patterns are in here? The model has to discover the structure without a teacher.
The most common form is clustering, which groups similar items together. Another is finding a simpler description of messy data, so that the important differences stand out. These methods are great for exploring data you do not understand yet, and for spotting anything unusual.
Unsupervised results need human interpretation. The model might find five customer groups, but it cannot tell you what they mean. One more note: the pretraining of large language models is often called self-supervised. The text itself provides the answers, since the next word is always known.
A news app groups thousands of incoming articles into topics that no editor ever defined, purely by noticing which stories use similar words.