Generative AI
Last Updated: July 29, 2026 | By Mihail Sebastian | AI Dictionary
AI models that create new content, such as text, images, audio, or code, by learning the statistical patterns of training data and sampling from them.
What is Generative AI?
Generative AI is a class of models that create new content, such as text, images, audio, or code, by learning the patterns in training data and producing new outputs that follow those patterns. It answers a different question from most machine learning: not “which category does this belong to?” but “what would a new example look like?”
Traditional models classify or predict from data that already exists. A generative model produces data of its own, which is why the same technology that drafts a contract can also fabricate a convincing fake.
Types of Generative AI Models
- Generative adversarial networks (GANs) pit two neural networks against each other: a generator produces synthetic samples and a discriminator judges them, until the fakes pass for real. Ian Goodfellow and colleagues introduced the approach in 2014.
- Variational autoencoders (VAEs) compress data into a latent space of probability distributions, then sample from that space to decode new examples.
- Transformer language models such as GPT generate text by predicting the next token over and over, each prediction conditioned on everything written so far.
- Diffusion models start from random noise and remove it step by step until an image emerges; they drive most current image generators.
Applications of Generative AI
- Content drafting: first drafts of text, code, designs, and translations that a human then edits and approves.
- Synthetic data: artificial training records for cases where real data is scarce, expensive, or privacy-restricted.
- Deepfakes: the same face- and voice-generation ability turned to impersonation and misinformation.
Example of Generative AI
ChatGPT, launched by OpenAI in November 2022, put generative AI in front of the general public. Ask it to explain quantum computing to a ten-year-old and it does not retrieve a stored answer. It converts your prompt into tokens, predicts the most probable next token, appends it, and repeats until the explanation is complete.
Every sentence is manufactured on the spot. That is also the technology’s central governance problem: a system built to produce plausible text produces plausible falsehoods with equal fluency, a failure known as hallucination. For a longer walkthrough of the mechanics, see how ChatGPT really works.
Related AI terms: GPT · Generative Adversarial Network · Variational Autoencoder · Multimodal AI
Did you like the Generative AI gist?
Learn about 250+ need-to-know artificial intelligence terms in the AI Dictionary.
Mihail Sebastian — Writes about AI governance, regulation, and the technology behind them. Placeholder bio — replace with a real credential line. About