Keep pulling the thread on White House.
All Figure robots are controlled by an internally designed vision-language-action neural network called Helix.
Figure has transitioned its robot control systems from being manually coded with hundreds of thousands of lines of C++ to being controlled purely by its Helix neural network.
Figure's robots run inference for the Helix control model on-board using GPUs, allowing them to operate without a continuous internet connection.
Figure has an internal project called Vulcan that enables its robots to remain standing and mobile even after losing power or communication to a joint in their legs.
Figure's primary bottleneck for achieving large-scale deployment is the need for more diverse, high-quality training data.
Figure plans to offer its in-home robots for a monthly fee of $400 to $600, similar to a car lease.
In 2023, Figure deployed its robots for six months on a BMW body shop factory line to assist in building cars.
Figure's leadership believes that artificial general intelligence (AGI) may be achieved first in embodied forms like humanoid robots due to the importance of real-world interaction data.
In March 2024, Figure produced more robots than in the entire previous history of the company combined.
Figure reduced the cost of its humanoid robot by approximately 90% between its Figure 2 and Figure 3 generations.
Figure's upcoming Figure 4 robot is expected to be the most significant generational leap in the company's history, described as its "iPhone one moment."
Figure's next-generation hand is designed to have a level of dexterity equivalent to a human hand, with a similar number of joints.