Article

Opinion: VLA models alone cannot achieve safe humanoid robot deployment

A Reddit user argues that vision-language-action (VLA) models alone are not enough to safely integrate humanoid robots into daily life. Humanoid robots far exceed humans in strength and speed; without complex multi-sensory input, learning from video by vision alone cannot master the human instinct to adjust force at the right moment in tasks such as holding a child's arm, carrying a cake, or closing a door, which can easily cause accidents and hinder the industry's adoption.

Part of the event: Opinion: VLA models alone cannot achieve safe deployment of humanoid robots

This page provides a summary and source link only.