This seems cool: up to 4.5kg grasp load (though very much depending on grasp), seems like it has passive/mechanical backdrivable joints, and active stiffness control. Seems much, much more like a human hand than most things I have seen. Implemented in part via a low-backlash, low-damping reducer; which is interesting. Potentially how they avoid the need for qdd/low gear ratio actuators, but still get mechanical transparency.

Zero-shot sim to real for Asimov. I do think this is an important achievement -- we often understate the degree to which having the whole international community working on Unitree G1 robots made sim to real "easy" for that robot

With the ICRA deadline only a few days away, it's a great time to look over my guide for how to fake a robotics paper: - Don’t let other people run comparisons. If other people can try your model out, they’ll quickly uncover its limitations. If you can keep your model secret, that’s best. - Control the environment of your demo carefully. Lighting, objects, initial robot configuration, and so on. This lets you overfit the demo scene and get really nice, smooth, high quality motions. - Never show your failures. This one might seem obvious — why would I show failures? — but if people can see where your model fails, they can start to see the limits of what you can do. Only strong robotics papers and results can show failures with confidence. - If you have to let other people run comparisons, choose the people carefully. Make sure they’re only happening in the right circumstances, ones your model supports as roughly in distribution. Absolutely don’t do what Physical Intelligence or NVIDIA do and open source your model so anyone can benchmark it. - When working on the results section of your research paper or blog post, you may be tempted to include some baselines. This is a good idea; just be careful to choose weak baselines so you look good. - On the same note: cherry-pick your benchmarks. There are a ton of robotics benchmarks out there, and they all test subtly different things. Importantly, these differences (obviously don't do these things please)

Robots need to be useful and need to be out in the world doing a variety of stuff to get better. They can't just learn in labs. In the end the two things that stick out are - a robot is a tool. It has to actually do a job, and if youre not trying to use it for that job it probably doesnt work - along the same lines, what you see is what you get -- when you see a demo you have to remember its the absolute best the robot has ever done, and if you didnt see it do something, you must assume it cant do that thing

In-Context Learning Results Hint at a “GPT Moment” for Robotics For general-purpose robots to be useful and economical, you need to be able to teach them new skills on the fly. And now we can see how this will work, with in-context long horizon video demonstrations shown by companies like Generalist, Skild, Rhoda, Rhoda, and RobbyAnt. The idea is simple: prompt with a demonstration, mixed video and proprioception information. Language isn't enough -- it doesn't specify the problem well enough. And it's starting to really work. more in blog post

A teleop competition is a fascinating artifact of the times. Building a really good teleop stack is in fact worthy of competition, the robot is a product after all and this is the user experience. World humanoid robot games seem incredibly well thought out

This is the mark of a general purpose robot foundation model -- that it can be taught tasks from examples without training. We saw an early version of this from rhoda and a solid one last week from generalist, and now a really impressive, ten minute long one from skild. This to me does feel like the true gpt moment for robotics -- because if it truly generalizes, that means that you can really start deploying robots for anything, as developing robot skills becomes like prompting

Really cool video but I actually hate how everyone is like "this is how china's robot packs itself" or "china's robot beat Usain bolt" -- i am sure there is a more useful description than the country that ships 97% of all humanoid robots, there are dozens of companies

I have believed for a few years now, that what the field badly needed was accessible, low cost, capable platforms so everyone could build and rapidly iterate i am very happy right now that we've got several companies building stuff like this, looks great

@chris_j_paxton “people dont actually want robots” ❌ “people dont actually want robots that do not work” ✅ We have given classical robotics too much time. learning will solve lower dynamics problems one by one