Self-supervised learning (SSL) approaches take unstructured/unlabeled data as input to the model. Unlike unsupervised learning, a SLL model annotates the data on its own, and labels that have been predicted with high confidence are used as ground truths in future iterations of the model training. Essentially, it makes use of the structure within the data to generate its own labels.
Typically, self-supervised learning identifies some secondary task where labels can be automatically obtained, and then trains the network to do well on the secondary task.
Perhaps it is incorrect to think of SLL and unsupervised learning as separate techniques. Self-supervised learning IS an unsupervised learning algorithm that uses certain methods to derive learning signals without explicit labels.