Cybernews Cybernews
19 days ago

The robotics industry might have found a highly efficient bypass for its biggest data bottleneck.

Reward AI recently introduced OM-1, a general-purpose robot manipulation policy trained entirely without traditional teleoperation or on-robot experience. The system learns directly from natural human demonstrations captured through a wearable sensor device known as the Omnibody Hand.

By processing tactile signals, visual context, and force measurements from this human data, the OM-1 robot foundation model translates movement into actions across various industrial arms and humanoid robots. The company reports that this approach to physical AI allows the system to learn new tasks from under 30 minutes of human demonstration.

What industry do you think will be disrupted first by robots that can learn physical tasks this quickly? Leave your thoughts in the comments.

Media: Reward AI
229

Comments 0

Replying to
5000 max

Be the first to comment!

This website uses cookies
This website uses cookies.
To maintain our services, we use cookies. They allow us to personalize ads, analyze our traffic, and provide social media features.

Customize your preferences

Select the cookies you accept: