What is "deep learning"?

Deep learning is a form of machine learning built on neural networks with many stacked layers, each layer transforming the data a bit further before passing it to the next. This depth lets the network learn increasingly abstract patterns, for instance, early layers in an image model might detect simple edges, while deeper layers combine those into shapes, then objects, then whole scenes, an approach that proved dramatically more capable than earlier, shallower techniques once enough data and computing power became available.

Studying while you sleep is a real, though different, concept from cognitive science with no connection to layered neural networks. Deleting deeply buried files is an ordinary computer file-management task, unrelated to how AI models learn patterns from data.

The breakthrough moment for deep learning in the wider public eye came when deep neural networks began sharply outperforming older techniques on image recognition contests, a result that helped trigger the massive wave of investment and research that led directly to today's large language models and image generators.