Loading video...
Video Failed to Load
Our first customer use case took 12 months, our second customer use case took just 30 days Helix learned high-rate logistics with a single neural network On Sunday, we successfully validated this on-site at the customer
477,323 views • 1 year ago •via X (Twitter)
11 Comments

Last week, Figure unveiled Helix: a vision-language-action model for human-like reasoning Today, we're applying Helix to a real-world application: high-rate, small package logistics This is a task previously unsolvable with traditional robotics techniques

What's exciting is that we accomplished this with our second customer use case in just 30 days Our first customer use case took us 12 months End-to-end learning approaches are allowing Figure's robots to learn new tasks quickly!

Today, we published a comprehensive write-up on our latest Helix advances If you're interested in an in-depth technical discussion, check out the report: → Implicit stereo vision → Multi-scale visual representation → Learned visual proprioception → Sport mode

Consider joining our Helix team to help scale Embodied AI to millions of robots: → Helix, Training Infra → Helix, Large Scale Training → Helix, Manipulation Engineer → Helix, Large Scale Model Evals → Helix, Reinforcement Learning

We're excited to continue pushing humanoid robots into the physical world If you're interested in following our journey, give us a follow: @Figure_robot

One AI platform powers the next generation of exceptional customer service

Wouldn’t it be more cost effective to modernize the sorting vs a very costly robot?

Figure 02s arriving at the facility

How was this "unsolvable"? To me it just looks like some conveyor belts and a camera to detect barcodes? Am I missing something?

I don’t think I’ll ever get tired of watching humanoid videos

doesn’t it make more sense just to stick with a giant sorting machine? we replaced human workers on logistic lines because they were slow and inaccurate, so I don’t see how replacing a dedicated sorting machine with humanoid robots is the right approach here - this doesnt seem like a human orientated activity



