What Is a Generative Adversarial Network (GAN)?

A generative adversarial network (GAN) is a machine learning architecture where two neural networks train against each other: a generator that creates fake data and a discriminator that tries to tell the fakes from real examples. As they compete, the generator gets steadily better at producing realistic output, such as images, audio, or synthetic data.

How a GAN works

The two networks play opposite roles. The generator starts from random input and tries to produce data that looks real. The discriminator is shown a mix of real examples and the generator's fakes, and it tries to label each one correctly. Every round, the discriminator gets better at catching fakes, which pushes the generator to make more convincing ones.

This back-and-forth is the "adversarial" part. Training ends when the generator's output is realistic enough that the discriminator can no longer reliably tell real from fake. At that point the generator can produce new, believable data on its own.

Why GANs matter for AI

GANs were a major step in generative AI. They showed that a model could learn to produce convincing new images and other media, and for years they set the standard for realistic image generation. They remain useful for generating synthetic data when real data is scarce or sensitive.

For newer image generation, diffusion models have largely taken over, because they are more stable to train and easier to control. GANs are still worth understanding as the architecture that proved the idea and shaped what came next.