Thanks to some high-profile tech billionaires, humanoid robots have suddenly become a hot topic, both in positive and negative ways. Seen as the culmination and combination of technologies that power robotic arms and robot dogs like Spot, these robots are sparking controversy and discussion about the future, especially when these machines become available for consumer purchase and use in homes. That’s the kind of future that Figure AI is trying to paint, and it just took one step closer to that impressive and unsettling scenario.
This scenario pretty much involves house robots cleaning up after their human owners, putting away not only toys but also groceries to place them in their proper storage or section in the refrigerator. The clincher? These robots may have never seen some of these objects before, but thanks to Helix AI, they’re able to not only identify what they are but even work together to finish the task at twice the speed.
Designer:
Imagine coming home after a tiring trip to the supermarket, carrying not one but two bags full of groceries. The last thing you probably want to do at that point is to put them away, no matter how important you know the task is. Maybe you want to just spend some time with your kids, or play with your pets, or even sit back and chill for a while. If you have two Figure robots and Helix AI, you can definitely do that without worrying about the groceries.
Helix AI is a Visual-Language-Action or VLA model, which is a very fancy term for AI that uses both visual and language processing to perform actions in the real, physical world. While many humanoid robots are demonstrated to be able to navigate and maneuver paths, do back flips, or play table tennis, Figure’s machines are being trained to do household chores, specifically picking up after humans.
There are a few things that make Helix sound pretty impressive, one of which is the AI’s ability to actually understand which object you’re referring to even without specifically naming it. Tell the robot to pick up the “desert item” and it will correctly identify a cactus toy from a collection of other objects. Helix also works on objects it has never seen before, using language context and other models to properly identify and classify them.
The biggest and most impressive demonstration of this AI model is its ability to coordinate work between two robots. Each robot is able to silently communicate which item goes where and hands them off to whichever robot is nearest to the drawer or the refrigerator. Of course, humans can do this faster and more efficiently, even if it means losing time or energy talking to each other. This is just the tip of the iceberg, of course, and the commercial robots of the future will probably move faster and look less inhuman.
The post first appeared on .
Related Posts
