Figure’s Helix AI Model Aims to Transform Home Robotics
Figure, the Bay Area robotics startup led by founder and CEO Brett Adcock, has unveiled a groundbreaking AI model designed to bring humanoid robots into homes. The new system, called Helix, is a Vision-Language-Action (VLA) model that enables robots to process visual and verbal commands in real time.
The announcement comes just weeks after Figure decided to move away from a collaboration with OpenAI, signaling a shift toward independent advancements in humanoid robotics.
How Helix Works: Bridging Vision and Language for Robotics
Helix operates similarly to Google DeepMind’s RT-2, which integrates video-based training with large language models (LLMs). By combining visual data and language processing, Helix enables robots to understand spoken instructions and execute tasks accordingly.
According to Figure, the model demonstrates remarkable object generalization, allowing robots to handle thousands of household items with different shapes, sizes, and materials—none of which were included in their initial training.
“In an ideal world, you could simply tell a robot to do something, and it would just do it,” Figure states. Helix is designed to turn that vision into reality, processing spoken instructions like:
- “Hand the bag of cookies to the robot on your right.”
- “Receive the bag of cookies from the robot on your left and place it in the open drawer.”
The model even allows multiple robots to coordinate tasks, enhancing efficiency in household settings.
Shifting the Focus to Home Robotics
Most humanoid robotics firms prioritize industrial applications, using warehouse and factory environments to refine their systems before considering home use. High costs and complex learning hurdles have kept household robots on the back burner.
However, Figure is taking a different approach. By focusing on home environments—where variables such as furniture arrangements, lighting, and object placement constantly change—Figure is embracing a more challenging, yet ultimately more rewarding, frontier for humanoid robotics.
Scaling AI for Adaptive Home Environments
A major hurdle in home robotics is scalability. Currently, teaching robots new behaviors requires either extensive manual programming or thousands of training repetitions. Figure notes that such methods are impractical for dynamic home settings, where adaptability is key.
“For robots to be useful in households, they will need to be capable of generating intelligent new behaviors on demand, especially for objects they’ve never seen before,” the company explains.
Helix aims to overcome these challenges by enabling robots to learn new tasks through interaction rather than pre-programming. This adaptability is crucial for tasks such as cooking, cleaning, and organizing—where every home presents a different set of challenges.
Early Stages, Big Ambitions
Despite the promising potential of Helix, Figure acknowledges that the technology is still in its early phases. The company is actively recruiting engineers to help develop the model further. While polished demo videos showcase Helix’s capabilities, much of the work remains behind the scenes.
Still, the announcement marks a significant step toward a future where humanoid robots can seamlessly integrate into daily life. If successful, Helix could redefine what home robotics can achieve, moving the industry closer to a world where robots don’t just exist in factories—but also in our kitchens, living rooms, and beyond.
