What is Generative AI?

Generative AI produces content, including text, images, music, and computer code. A request might produce a first draft of a letter, an illustration, or a proposed software function. The output gives a person something to examine and revise.

A request and context enter a trained generative model, which produces new content for human review.
Generative models produce new content from learned patterns and the current request. Different models support different media; the result still needs review.

Think of a writer preparing several drafts from the same brief. A generative model can offer different versions, drawing on patterns learned during training. You still have to decide whether any of them serves the purpose.

How Does Generative AI Work?

At its core, Generative AI operates using algorithms that are able to learn from a set of data and then generate new data similar to the original set. These algorithms are a part of machine learning models called Generative models, like Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs).

An image-generation model learns patterns from training images and uses those patterns to produce images. A generated result can still resemble material from training; it should not be assumed to be entirely unique. The NIST Generative AI Profile identifies risks involving memorized training material in generated output.

Real-world Applications of Generative AI

Generative AI isn't just a theoretical concept – it has practical applications across a wide range of industries. Here are some generative AI examples:

  • Art and Design: Generative AI can help create images and design concepts from a brief. Image-generation examples include OpenAI’s DALL-E 2 and Midjourney. Review the result against the intended purpose before using it.
  • Writing and Journalism: Generative AI can help draft articles, blog posts, or reports, producing a first pass that human writers can check, refine, and polish. A useful starting prompt specifies the audience, purpose, source material, and desired length.
  • Music and Entertainment: Generative AI algorithms can create new pieces of music or script dialogues for a movie, offering a new tool for musicians and filmmakers.
  • Healthcare and Medicine: Generative AI applications include producing synthetic patient data for research and proposing molecules for drug discovery.

The Future of Generative AI

The useful question is what a generated result contributes to the work. Does it save drafting time? Offer an idea worth developing? Introduce errors that take longer to correct? Those comparisons help determine where generative AI is worth using.

The guides on this site explain prompts, context windows, and other concepts that help you make those comparisons.