Unsupervised learning
ML where the data isn't labelled — the model finds patterns and structure on its own.
Clustering (group customers), dimensionality reduction (compress data), embeddings (represent words/images as vectors). Pre-training LLMs is largely unsupervised on raw text.