We fuel robots with real-world data.

Unlike the vast amounts of internet data used to train ChatGPT, real-world data for robotics applications—such as grasping and manipulation—is extremely scarce yet highly valuable. The ratio of internet data to real-world robotics data is roughly a million to one.

This is a bottleneck in robotics research. Researchers often rely on simulation data because it's easy to generate, but policies trained in simulation face a reality gap when applied to real-world scenarios.

BaronX will fuel robotics and AI research with this data, accelerating progress towards generally intelligent robots.

Team:

Jack Armstrong

Raymond Lee