Self-Supervised Learning
A machine learning paradigm where a model automatically generates its own training labels from the raw input data by hiding and predicting parts of it.
Think of It Like This
Like a student studying for an exam by covering up random words in their textbook and guessing what they are.
This approach bypasses the massive bottleneck of manual human labeling. Models learn rich representations of text, images, or audio by predicting missing words (BERT), next words (GPT), or matching augmented image pairs (SimCLR). It is the engine that powers the creation of all modern foundation models from vast uncurated internet data.