Trending storyDeveloping

Latent.Space interview explains world-action models for robot control

1 article1 sourcesince Oct 13Last article 1h ago · Oct 13, 1:42 PM ET

Overview

AISummary of 1 article

Latent.Space announces a podcast episode with Jason Ma of Dyna Robotics on world-action models, which predict future states to guide robot actions.

The episode covers training on over a million hours of human video. Ma says this lets robots adapt to new tasks with as little as 45 minutes of demonstrations.

Written by AI from the articles below · updated Oct 13, 1:45 PM ET

Check the sources:

Latest developmentOct 13, 1:42 PM ET

Latent.Space interview explains world-action models for robot control

Article timeline

The articles in this story. Times are ET.

Oct 13

1:42 PM
Latent.Space@latentspacepodX
Latent.Space interview explains world-action models for robot control
AILatent.Space announces a podcast episode with Dyna Robotics' Jason Ma, who explains how world-action models predict the future to guide robot actions. The episode also covers training on over a million hours of human video, which lets robots adapt to new tasks with as little as 45 minutes of demonstrations.

Heat trend

Not enough continuous observations to show a trend yet.