Also known as: GenAI, Gen AI
In plain English
Most older AI sorted or predicted things. Generative AI makes new things: it can write an email, draw a picture or compose music when you describe what you want.
In practice
Common business uses are drafting documents, summarising meetings, writing code and answering customer questions. The main risks are confident errors (hallucinations), leaking sensitive data in prompts, and copyright questions over generated content.
Under the hood
Generative models learn the probability distribution of their training data and sample from it. Text models are usually autoregressive transformers that predict the next token; image models are often diffusion models that learn to remove noise step by step.
Example
"Marketing used generative AI to draft 20 versions of the campaign email."