Glossary · Term

vision-language-action model

← all terms

Definition

Plain language

An AI that takes in camera images and an instruction and directly outputs robot motor commands.

As stated in the literature

VLA — a model mapping pixels plus a natural-language instruction straight to low-level robot actions, end to end; contrasted with code-writing agents that compose modular tool calls.

Also called: VLA, VLAs, vision-language-action models

Why it matters: By going straight from images and words to motor commands, it skips hand-built modules, but that end-to-end style can be brittle on unfamiliar tasks.

For example, shown a cluttered desk and told 'hand me the red mug,' it outputs the arm movements to do it directly.

Heard on the show

“For contrast, the end-to-end vision-language-action models — the "one big network maps pixels to actions" approach — mostly collapse here.”
Episode 194 — How a Robot Builds a Debugging Notebook It Can Read, Edit, and Hand to Another Robot

Mentioned in 2 episodes

  1. 194
    How a Robot Builds a Debugging Notebook It Can Read, Edit, and Hand to Another Robot
  2. 161
    A Robot That Plays Before You Give It a Job, And Why That Beats Retrying

Related terms