Field
Generative AI
Systems that make things — images, and increasingly everything else.
Generative AI produces new content rather than classifying existing content. The distinction sounds academic and isn't: a classifier tells you what's in a photo, a generative model makes the photo.
This field covers the two approaches that mattered. GANs came first — a generator and a discriminator locked in a contest — and dominated image generation for years before their training instability caught up with them. Diffusion models largely replaced them, and are the reason image generation ended up on consumer hardware and in public hands.
Multimodal AI is where this is heading: models that take images, text and audio in one shared representation. They're strong at describing a page and weak at reading it precisely, which explains most disappointment with them.
Start with Diffusion Model — it's what's actually running when you generate an image.
3 concepts in this field
Diffusion Model
How most AI image tools work — starting from random noise and removing it step by step, guided by a prompt, until a picture appears.
GAN (Generative Adversarial Network)
Two networks trained against each other — one faking, one detecting — until the fakes pass. The technique diffusion largely replaced.
Multimodal AI
Models that handle more than one kind of input — text and images, sometimes audio and video — in a single shared representation.