Figure Introduces Helix, a Vision-Language-Action Model for Humanoid Robots

Figure introduced Helix, a vision-language-action AI model that gives its humanoid robots full upper-body control of wrists, fingers, torso and head, directed by natural-language commands. Helix was trained on roughly 500 hours of teleoperated demonstration data, with its low-level motor-control policy pretrained in simulation, and it runs entirely on the robot's onboard GPUs rather than a remote server. Figure said the model lets its robots pick up thousands of previously unseen household objects on command, a step toward general-purpose manipulation rather than tasks scripted in advance. Figure AI is a Parkway Venture Capital portfolio company.

Read the full article: via Figure AI