Figure’s humanoid robot will do your chores with voice commands

We've seen robots run, jump, and even dance their way into the future

But what about robots that could do your chores with a simple voice command

Robotics company Figger just released Helix, a new proprietary AI system that allows its robots to complete complex tasks from verbal commands

No pre-programming, no training, just action. Many humanoid robots today need to be trained through demonstrations

in order to perform certain tasks like picking up certain objects or programmed with exact instructions

Helix, on the other hand, is designed to learn and adapt in real time

making robots more practical for everyday use. The model can also operate simultaneously on two robots

allowing them to work together on tasks involving new objects they've never seen before

In Figure's latest demo video, two robots are asked to put away a bag of groceries

without specific instructions on how or where to store the items. From that single instruction

the robots are able to scan the kitchen, take the items out, and put them into the fridge

all while working together throughout the process, handing items off to one another

So, how do they pull this off? Helix is a VLA model which stands for vision action language The robots visualize the environment around them interpret language commands and then act on them Figure describes it as a first of its kind System 1 System 2 VLA model

which basically means Helix uses two different systems to control the robot's upper body

According to Figer, previous systems could either handle many tasks slowly or perform a specific task quickly

Helix combines both approaches in order for the robots to be faster and more adaptable

Helix is still in its early stages, but Figer is moving fast, planning to begin home alpha testing in 2025, two years ahead of schedule

And while robot roommates aren't here just yet, Figger also announced a new real-world application for its robots, handling and sorting packages

The announcement comes with an update to the VLA model, which includes better depth perception for more accurate movements, improved vision processing to recognize both small details and the overall scene, making objects easier to handle, and self-calibration

so robots can adjust themselves automatically, improving reliability and consistency. This could be a big step towards making robots more useful in factory and warehouse settings

and is a reminder that while we're still living through the early days of humanoid robotics

progress is happening fast