https://www.kadenze.com/courses/creative-applications-of-deep-learning-with-tensorflow-iv/sessions/unsupervised-and-supervised-learning
Autoencoders are a type of unsupervised machine learning. It's a specialised type of neural network that aims to take high dimensional data, and re-map that to a new representation with lower dimensionality.
Supervised methods of machine learning involve significant amounts of human labour to annotate, prepare and define representations of data to feed into learning algorithms . Automating this process is a currently an active field of machine learning research aimed at removing this requirement.