Supervised learning uses labeled data to train models for predicting outcomes like spam detection, while unsupervised learning works with unlabeled data to discover hidden patterns like customer segmentation.