Figure Launches Project Go-Big to Build World’s Largest Humanoid Training Dataset

• Project Go-Big captures large-scale, real-world human video to train Helix, Figure’s Vision-Language-Action model.
• Helix now navigates from speech using zero-shot transfer, with no robot demonstrations or task-specific training.

Figure has launched Project Go-Big, a new effort to create the largest and most diverse dataset ever used to train humanoid robots. The goal is to teach Helix—Figure’s Vision-Language-Action model—how people move and act in everyday life by using first-person video recorded in real homes. Instead of relying on simulations or choreographed tasks, the project uses real human behavior as the foundation for training.

"The biggest breakthroughs in machine learning have come from pretraining large neural networks on massive, diverse datasets: ImageNet for vision, Wikipedia for language, or YouTube for generative video models. Unlike vision or language, robotics lacks a large-scale equivalent—no "YouTube for robot behaviors," the company posted on its official blog.

To make this possible, Figure is collecting video across a wide range of real environments, with major support from Brookfield. With access to over 100,000 residential units and millions of square feet of commercial space, Brookfield gives the project scale and variety that would be hard to match otherwise. This helps Helix learn from the kinds of spaces and behaviors it would actually see in the real world.

So far, Project Go-Big has already delivered. Helix can now move through real homes by following voice commands like “go to the fridge” or “walk to the kitchen table.” The model learned this entirely from human video, with no robot demonstrations needed. In addition, one network now handles both movement and object handling, marking what Figure says is the first time a humanoid has learned navigation and manipulation end-to-end using only video of people.

🌀 Tom’s Take:

Project Go-Big shows that Figure is serious about leading in humanoid robotics. Helix is their big bet on building robot intelligence through real-world data.

Source: Figure