What is unsupervised ML?

In books and courses about ML, it is often explained that there are three different kinds: supervised learning, unsupervised learning, and reinforcement learning. Sometimes, combinations will be explained, such as semi-supervised learning. With supervised learning, we provide data (X) and an answer (y), and the model learns to make predictions. With unsupervised learning, we provide data (X), but no answer (y) is given. The goal is for the model to learn to identify patterns and characteristics of the data by itself, and then we use those patterns and characteristics for something else. For instance, we can use unsupervised ML to automatically learn the characteristics of a graph and convert those characteristics into embeddings that we can use in supervised ML prediction tasks. In this situation, an unsupervised ML algorithm is given a graph (G), and it generates embeddings that will serve as the training data (X) that will be used to be able to predict answers...