All terms

Generative AI

AI systems that create new content — text, images, audio, code — rather than only classifying or scoring inputs.

AI Foundations1 min read

Definition

Generative AI models learn a data distribution and sample new examples from it: next-token prediction for text, diffusion for images, and related methods for audio and video.

It contrasts with purely discriminative models that only output labels or scores.

In simple terms

A discriminative model is a food critic (scores dishes). A generative model is a chef (creates new dishes in a learned style).

Where you see it

  • ChatGPT-style assistants and code copilots.
  • Image generators and voice synthesis.
  • Synthetic data for augmenting training sets.

How it works

  1. 1.Train on large corpora

    Learn patterns of the modality.

  2. 2.Condition on prompts

    Steer generation with text or other inputs.

  3. 3.Sample outputs

    Decode tokens or denoise until a finished artifact.

Why it matters

  • Generative AI is the product category reshaping creative work, software, and language tech.

Often confused

  • Generative AI always tells the truth.

    It generates plausible content — facts still need verification (see hallucination).