Deep learning is a subfield of machine learning that uses neural networks with many layers (thus “deep”) to learn representations of data. It is characterized by the use of multiple layers of nonlinear processing units, each transforming the data (starting from raw input) into higher-level abstractions. For example, in image recognition, early layers detect edges, mid layers detect patterns like corners or textures, and later layers detect object parts and then whole objects. Deep learning gained prominence with the success of deep convolutional networks in vision and LSTMs/transformers in NLP, fueled by large datasets and powerful GPUs. It’s essentially representation learning with neural networks, often requiring large amounts of data and compute, but yielding state-of-the-art results in many tasks. Architectures include CNNs, RNNs, GANs, Transformers, etc., each suited for different data modalities and problems.