In the field of robotics technology, American startup Figure AI recently released a big end-side model called Helix, marking a major breakthrough in humanoid robot control technology. Helix is the first visual-language-action (VLA) model that enables high frequency, continuous control of the entire humanoid upper body (including the head, torso, wrist and finger). This means that robots can perform tasks directly based on natural language instructions without a lot of training.
Figure AI solves the contradiction between universality and speed of visual language models by establishing two complementary systems. System One is a fast-reacting visual motion strategy model that can perform 200 precise actions per second, while System Two is an open source visual language model that can understand complex scenarios and natural language. These two systems enable efficient communication through end-to-end training, thus overcoming the limitations faced by humanoid robots in performing diverse tasks.
What is even more exciting is that Helix can not only run on a single robot, but also is the first AI model that can achieve dual-machine collaboration. In a series of demonstrations, two robots equipped with Helix successfully collaborated on a collaborative classification against a bunch of unseen debris. In order to complete the task, the robots need to make visual "confirmation" and look at each other to ensure accurate information transmission. This way of human-machine collaboration not only demonstrates Helix's powerful capabilities, but also implies that the interaction between humanoid robots and humans will be more natural and efficient in the future.
With just 500 hours of high-quality supervision of data, Helix can easily identify and process thousands of small household items. With simple instructions, the robot can accurately complete tasks that have never been demonstrated, showing its amazing flexibility and intelligence.
With the launch of Helix, Figure AI is also actively raising funds, with a target of $1.5 billion, and the company is expected to value it at $39.5 billion. This number has increased by nearly 14 times compared with last year, reflecting the market's huge interest and potential in robotics.