Unsupervised machine learning is a branch of machine learning that studies learning techniques that autonomously detect patterns in data. Unlike supervised machine learning, only the raw observations are provided, so there is no need to create examples for the model to learn from.
Unsupervised learning is useful to find patterns of the real system that produced the data, but the patterns are inherent to the dataset used to train the model. They are not, in general, extensible to other datasets. As such, unsupervised techniques (and their models) are much less universal than supervised techniques.