Decoder. plain-English AI glossary

Generative Adversarial Network

▼ Fading

Also called GAN

The full name for a GAN — a generator and a critic trained by competing against each other.

Think of it like

Counterfeiter versus bank teller: the better the teller gets at spotting fakes, the better the counterfeiter has to be.

Example

Ian Goodfellow’s 2014 paper introduced the idea and set off years of photorealistic-face research.

How it actually works

The generator maps random noise to fake data; the discriminator learns to separate fakes from real samples. Their opposing objectives create a moving target that, when it works, produces sharp samples — and when it doesn’t, collapses or oscillates. The adversarial framing was the big idea; the training pain was the price.

For product teams

Historically important, but you’ll rarely reach for it in a new product today.

For engineers

Same as GAN: adversarial training of paired generator and discriminator networks.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome