Chris Paxton's banner

Chris Paxton

@chris_j_paxton • 37,704 subscribers

Mostly posting about robots. currently AI @agilityrobotics prev embodied AI @AIatMeta, @NVIDIAAI. All views my own.

Shorts

Always important to remember that a lot of these robots are "faking" the humanlike motions -- its a property of how they're trained not an inherent property of the hardware. They're actually capable of way weirder stuff and way faster motions.

Always important to remember that a lot of these robots are "faking" the humanlike motions -- its a property of how they're trained not an inherent property of the hardware. They're actually capable of way weirder stuff and way faster motions.

9,873,060 Aufrufe

This is looking amazing

This is looking amazing

254,823 Aufrufe

Impressive reaction time from the new Sharpa robot at IROS. Catching randomly falling foam cylinders. I did not do nearly this well

Impressive reaction time from the new Sharpa robot at IROS. Catching randomly falling foam cylinders. I did not do nearly this well

13,375 Aufrufe

Berry pickup from Kyber Labs Robot hands that can pick up fragile objects quickly will be important to solving long-tail of robotics problems not currently automated

Berry pickup from Kyber Labs Robot hands that can pick up fragile objects quickly will be important to solving long-tail of robotics problems not currently automated

33,336 Aufrufe

Using monocular slam + ego video to create usable reinforcement learning environments for embodied ai

Using monocular slam + ego video to create usable reinforcement learning environments for embodied ai

12,943 Aufrufe

Predicting how the environment changes over time is a key task in robotics, enabling all kinds of downstream tasks. Trace Anything is a very interesting look at how you can train models which predict the trajectory each point will take in the future -- you can use this for robot manipulation, video prediction, and more. Learn more ->

Predicting how the environment changes over time is a key task in robotics, enabling all kinds of downstream tasks. Trace Anything is a very interesting look at how you can train models which predict the trajectory each point will take in the future -- you can use this for robot manipulation, video prediction, and more. Learn more ->

58,763 Aufrufe

It seems so simple to humans: "First watch the video with a red border, then pick up and place the same cube twice." And yet it's very difficult for policies as they are now. Learn more about benchmarking for robot memory ->

It seems so simple to humans: "First watch the video with a red border, then pick up and place the same cube twice." And yet it's very difficult for policies as they are now. Learn more about benchmarking for robot memory ->

15,507 Aufrufe

A favorite of the new robot designs from Fourier -- the gr3 is soft(ish) and friendly

A favorite of the new robot designs from Fourier -- the gr3 is soft(ish) and friendly

44,042 Aufrufe

Pantograph robot building with jenga blocks. Focusing on RL is a great and fairly unique strategy

Pantograph robot building with jenga blocks. Focusing on RL is a great and fairly unique strategy

30,546 Aufrufe

Watching the agibot x2 at the hackathon yesterday. I think there are a few extra degrees of freedom over the unitree g1 and you can really see it in the motion

Watching the agibot x2 at the hackathon yesterday. I think there are a few extra degrees of freedom over the unitree g1 and you can really see it in the motion

22,281 Aufrufe

These crazy videos of Chinese delivery robots from Neolix (and maybe similar companies) are my new favorite thing

These crazy videos of Chinese delivery robots from Neolix (and maybe similar companies) are my new favorite thing

19,236 Aufrufe

Some random thoughts reading through the new RobbyAnt VLA paper: - 20,000 hours of data across 9 robots!! - damn, Chinese companies are going to trivially outscale the American ones on real robot data - It's really cool they train on depth; it means they can handle transparent objects really well for example - You can never tell how good these models are without trying them, since everyone trains on different robots, but from the results they show it does seem to clean up - cross embodiment scaling laws are really cool - if you have lots of robots do you need human video??

Some random thoughts reading through the new RobbyAnt VLA paper: - 20,000 hours of data across 9 robots!! - damn, Chinese companies are going to trivially outscale the American ones on real robot data - It's really cool they train on depth; it means they can handle transparent objects really well for example - You can never tell how good these models are without trying them, since everyone trains on different robots, but from the results they show it does seem to clean up - cross embodiment scaling laws are really cool - if you have lots of robots do you need human video??

19,921 Aufrufe

ManiFeel is a benchmark for contact-rich manipulation tasks. Force- and tactile-based tasks with occlusions and partially-observable state are going to be crucial for true general purpose robotics. Website:

ManiFeel is a benchmark for contact-rich manipulation tasks. Force- and tactile-based tasks with occlusions and partially-observable state are going to be crucial for true general purpose robotics. Website:

12,063 Aufrufe

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

Cotton picking robot with 108 arms

Cotton picking robot with 108 arms

6,229,042 Aufrufe • vor 2 Monaten

Worked out of the box, first time. E2E mobile manipulation on Agility Digit v4 at #IROS2026

Worked out of the box, first time. E2E mobile manipulation on Agility Digit v4 at #IROS2026

28,784 Aufrufe • vor 22 Tagen

1x neo gamma vacuuming at GTC

1x neo gamma vacuuming at GTC

516,108 Aufrufe • vor 1 Jahr

The best humanoid robotics in the world from 10 years ago at the Darpa grand challenge. Its amazing how fast this has changed; imagine what this will be like in 10 years

The best humanoid robotics in the world from 10 years ago at the Darpa grand challenge. Its amazing how fast this has changed; imagine what this will be like in 10 years

291,368 Aufrufe • vor 10 Monaten

Easily the most impressive uncut autonomous video I've ever seen, from Physical Intelligence

Easily the most impressive uncut autonomous video I've ever seen, from Physical Intelligence

467,692 Aufrufe • vor 1 Jahr

Booster vs unitree dance off

Booster vs unitree dance off

17,178 Aufrufe • vor 21 Tagen

End-to-end whole body mobile manipulation on Agility Digit. We ran this demo for ~12 hours at IROS earlier this week video is 1x speed, autonomous, in a previously-unseen location (not in training data).

End-to-end whole body mobile manipulation on Agility Digit. We ran this demo for ~12 hours at IROS earlier this week video is 1x speed, autonomous, in a previously-unseen location (not in training data).

13,684 Aufrufe • vor 16 Tagen

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.

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.

20,012 Aufrufe • vor 27 Tagen

This is the first time that I have seen Skild doing live demos in public! Demonstrating neural nets for precision manufacturing

This is the first time that I have seen Skild doing live demos in public! Demonstrating neural nets for precision manufacturing

108,574 Aufrufe • vor 7 Monaten

Reach the banana hanging from the ceiling

Reach the banana hanging from the ceiling

69,130 Aufrufe • vor 4 Monaten

Galbot is a good dancer

Galbot is a good dancer

12,920 Aufrufe • vor 19 Tagen

Still one of my favorite robotics use cases: Peanut Robotics cleaning hotels and workplaces autonomously. Need a home version of this guy

Still one of my favorite robotics use cases: Peanut Robotics cleaning hotels and workplaces autonomously. Need a home version of this guy

236,087 Aufrufe • vor 2 Jahren

It is really incredible what these robots can do now from omnixtreme - new RoboPapers episode tomorrow

It is really incredible what these robots can do now from omnixtreme - new RoboPapers episode tomorrow

62,576 Aufrufe • vor 5 Monaten

Its funny how even with the wonderfully dynamic unitree g1, these actions are slow. Perception-> action and the embodiment gap are the big things slowing down robotics. This is the part of the video where they start manipulating stuff

Its funny how even with the wonderfully dynamic unitree g1, these actions are slow. Perception-> action and the embodiment gap are the big things slowing down robotics. This is the part of the video where they start manipulating stuff

94,514 Aufrufe • vor 11 Monaten

One of my favorite works in the "robot learning from human videos" space that I have seen lately. This is because it offers a really coherent and unique way to represent robot skills without a lot of the downsides we see from estimation-based approaches.

One of my favorite works in the "robot learning from human videos" space that I have seen lately. This is because it offers a really coherent and unique way to represent robot skills without a lot of the downsides we see from estimation-based approaches.

78,419 Aufrufe • vor 1 Jahr

The best way to get robust, high-quality robot performance is through reinforcement learning; but RL in either the real world or a traditional simulation has lots of limitations. Instead, Jiazhi Yang in RISE does RL in a compositional world model. Learn more ->

The best way to get robust, high-quality robot performance is through reinforcement learning; but RL in either the real world or a traditional simulation has lots of limitations. Instead, Jiazhi Yang in RISE does RL in a compositional world model. Learn more ->

34,102 Aufrufe • vor 4 Monaten

Zero-shot navigation of complex terrain. The true advantages of humanoid robots are their ability to work in the true difficulty of human environments. All using onboard sensing and compute. Take a look ->

Zero-shot navigation of complex terrain. The true advantages of humanoid robots are their ability to work in the true difficulty of human environments. All using onboard sensing and compute. Take a look ->

10,945 Aufrufe • vor 1 Monat

If you're wondering who these guys are, former Tsinghua/BIGAI team

If you're wondering who these guys are, former Tsinghua/BIGAI team

10,226 Aufrufe • vor 1 Monat

The future of robot butlers starts with mobile manipulation. We’re announcing the NeurIPS 2023 Open-Vocabulary Mobile Manipulation Challenge! - Full robot stack ✅ - Parallel sim and real evaluation ✅ - No robot required ✅👀

The future of robot butlers starts with mobile manipulation. We’re announcing the NeurIPS 2023 Open-Vocabulary Mobile Manipulation Challenge! - Full robot stack ✅ - Parallel sim and real evaluation ✅ - No robot required ✅👀

179,087 Aufrufe • vor 3 Jahren

This is what I want to see from a hand video, damn

This is what I want to see from a hand video, damn

50,329 Aufrufe • vor 8 Monaten