Neural Network

Overview

A neural network is the fundamental structure behind modern AI. It is made of simple units, often called neurons, arranged in layers and joined by connections, each carrying a weight that determines its influence. Data enters at one end, passes through the layers being transformed at each step, and emerges as an output. What makes the network powerful is that the weights are not set by hand but learned from data, allowing the system to capture patterns no programmer could specify directly.

How it works

Each unit takes in numbers from the previous layer, combines them according to its connection weights, applies a simple non-linear function, and passes the result onward. Stacking many such layers lets the network build up increasingly abstract representations of its input. Learning happens by feeding the network examples, measuring how wrong its outputs are, and using backpropagation to adjust every weight slightly in the direction that reduces the error.

Why it matters

Nearly every contemporary AI capability, language understanding, image recognition, speech, generation, is a neural network at its core. Their defining strength is learning the relevant features directly from data rather than relying on hand-crafted rules, which is what let them succeed on problems earlier methods could not. Their appetite for parallel computation is also why they and the GPU rose together.