GeneralistAI Releases GEN-1 Model for Robot End Effectors

Thousands of kinds of cheap, task-specific hands are far more practical for today’s models than expensive, fragile, 20+ DoF 5-finger hands. This demonstrates you actually don’t sacrifice useful dexterity. https://twitter.com/generalistai/status/2080292438057373947

Super cool! Tool use is one of the grand challenges in physical intelligence / reasoning. Seems the solution is just … train on more tools! I also like the compliance mechanism that’s made visible to the camera … https://twitter.com/GeneralistAI/status/2080292438057373947

💪Force is the language of the physical world Let alone multi-end-effector behaviors, impressive is the GEN-1 peeling the tape off the table. Its demonstrating an understanding of the force. Such a reasoning is refreshing, and very different from what's needed for contact-rich… Show more

It is always fascinating to watch the Generalist demos. The smoothness and precision make these complex tasks look effortlessly simple, but I can truly appreciate the immense engineering effort behind such elegant execution.… Show more

This is a good point. The orange part is the "force sensor," it's deformable and visible to the camera. Not actually an instrumented force sensor of course. Nothing but vision as far as I can tell https://twitter.com/JacobZietek/status/2080354043889877493

Really cool results. I feel like multipurpose hands *feel* more scalable but in the near term reliability is a real constraint for deployment https://twitter.com/GeneralistAI/status/2080292438057373947