Figure unveils first-of-its-kind brain for humanoid robots

California’s Figure has launched Helix, a groundbreaking Vision-Language-Action model that integrates perception, language, and control to advance robotics. Founder Brett Adcock claims Helix "thinks like a human," generalizing to most household items and requiring no prior training.

Helix enables nuanced upper-body manipulation, robust multi-robot collaboration, and interaction using natural language prompts. It employs a single set of neural network weights for diverse tasks, negating the need for extensive fine-tuning.

Operating on low-power GPUs, Helix is suited for commercial use and addresses challenges like speed and generalization.