Generative AI47

Turning a few photos or a line of text into a 3D shape

The open questions of rights around AI training and AI output

Pictures made with generative tools, and the debates around them

An invisible mark quietly embedded in what AI generates

A structure that learns by shrinking data down and rebuilding it

Erasing a photo's background and keeping just the subject

Adding fitting color to a black-and-white picture

Keeping the same subject looking the same across every shot

A device that pins an image's shape down with a rough sketch

Fake video or audio where AI copies a real face or voice

Removing only the flecks mixed in, leaving the original untouched

Peeling away blur, layer by layer, until a picture appears

A maker and a checker compete, and both get better

A 3D scene held as millions of scattered color blobs

AI that invents a result that never existed before

The dial that decides how literally a request gets followed

Looking at a picture and writing what's in it as a sentence

Building a new picture starting from a given one

Erasing part of a picture and filling it back in

Filling the space between two points with several in-between steps

The place where compressed values end up sitting

Fitting mouth shapes to the beat of a sound

Leaving the main body alone and training only a small add-on

When the generator settles on one winning output and just repeats it

Turning body movement into a record of point locations

Turning a few lines of request into a finished piece of music

The field for writing what shouldn't appear in the picture

A 3D scene that answers with light for any spot and direction

The blur mixed in — and also where generation starts

Widening a picture's frame and drawing the new edges to match

The yardstick for whether a result matches what was requested

Shaping an instruction so the result matches what you want

The number of dots a picture is made of

The number of passes it takes to turn blur into a picture

The starting number that reproduces the same picture again

Keeping a picture's content and swapping only its surface look

Practice data made up in place of the real thing

Building text by looking ahead and adding the next piece

Making a picture out of what's written down

Turning text or a photo into a moving scene

The technology that reads text out loud in a voice

Shrinks an image down, then grows it back while recovering position

Enlarging a picture by inventing the detail it never had

Slightly different results pulled from the same request

An autoencoder that remembers things as a range, not a point

The part that turns a sound blueprint into an actual waveform

A model of how the world works that you can run before acting