GAN
Generative Adversarial Network
ELI5 — The Vibe Check
A GAN is two neural networks fighting each other. One (the Generator) tries to create fake images that look real. The other (the Discriminator) tries to catch the fakes. They train together, each getting better because of the other. The result is a Generator so good it can create hyper-realistic faces of people who don't exist. Deepfakes are largely a GAN problem.
Real Talk
A Generative Adversarial Network consists of two competing neural networks: a Generator that synthesizes fake data and a Discriminator that distinguishes real from fake. Trained adversarially, the Generator learns to produce increasingly realistic outputs. GANs dominated image synthesis before diffusion models overtook them in quality and stability.
When You'll Hear This
"GANs were the go-to for image generation before diffusion." / "The deepfake was generated with a GAN."
Related Terms
Diffusion Model
Diffusion models generate images by learning to reverse noise. In training, you take an image and slowly add random noise until it's pure static.
Generative AI
Generative AI is AI that creates new stuff — text, images, code, music, video — rather than just classifying or predicting. ChatGPT writes essays.
Neural Network
A neural network is a system loosely inspired by the human brain — lots of little math nodes connected together, passing numbers to each other.
Training
Training is the long, expensive process where an AI learns from data.