Generative AI


Refers to a type of artificial intelligence that involves content creation from training data and predictive models. Content is created when a prompt is entered. The output— which might be an image, music, text, code, or another form of content—is generated based on a corpus of other work. 

The accuracy of generative AI is dependent upon massive troves of training data from diverse sources. Many ethical questions about AI involve how data sets are gathered and cleaned, and biases that might emerge through these methods.

What is generative AI in simple terms?

Software, trained to make predictions from a corpus of data, generates content in response to requests made by a user.

What are generative AI examples?

Some popular generative AI programs include ChatGPT, Midjourney, LaMDA, Bard, and Stable Diffusion.

How does generative AI work?

Imagine a text-based generative AI program that receives the prompt, “diner menu written in the style of F. Scott Fitzgerald.”

The program would deliver output, which draws from a corpus of the novelist’s work and from a corpus of diner menus—many of which might be available on the web—to predict how text in this style might read.

You might see something like a plate of bacon and eggs described as “big as the Ritz,” as these two distinct styles of text have been programmatically synthesized by the generative AI application.  

What is the difference between generative and general AI

Artificial intelligence, in a general sense, describes all kinds of autonomous technology. It includes physical computing, such as robotics and autonomous vehicles, as well as screen-based or software-based autonomous technology.

The output of generative AI, however, is content—music, text, video, code, etc—generated from a corpus of content.