正在加载视频...
视频加载失败
Today, we present a step-change in robotic AI Sunday. Introducing ACT-1: A frontier robot foundation model trained on zero robot data. - Ultra long-horizon tasks - Zero-shot generalization - Advanced dexterity 🧵->
2,058,122 次观看 • 10 个月前 •via X (Twitter)
67 条评论
Instead of teleoperation, we train solely on data from our Skill Capture Glove. The glove is co-designed with Memo's hand, meaning they share the exact same geometry and sensor suite. If you can do it wearing the glove, Memo can learn it.
The Skill Capture Glove gives us two orders of magnitude higher capital efficiency compared to teleoperation ($200 vs $20,000) It also allows us to scale diversity faster. You can collect data anywhere without needing to move robots around.
Skill Capture Glove aligns the hands, but what about the rest of the body? Human collectors vary in height and arm length, and are also visually different. We developed Skill Transform, a method that converts glove data into equivalent robot data with a 90%+ success rate.
It took us over a year to engineer the core infrastructure. We then spent the past 3 months to produce all the autonomous results above. Below, I highlight some of my favorite parts of this release.
The table-to-dishwasher task is the classic nightmare scenario for roboticists: Long-horizon, highly dexterous, precise, whole-body manipulation combined with delicate, transparent, reflective, and deformable objects. Yet Memo handles it so naturally and elegantly.
Specifically, wine glass loading is the most delicate subtask: Push down with too much force? Shatter. Insert the wrong prong? Shatter. We broke many during development, but zero over 20+ live demo sessions.
One less-known fact about glove-based data collection: it produces higher quality data than teleop on contact-rich tasks. Remote teleop can’t provide good force feedback, but gloves do naturally, making tasks like sock folding, which rely on feel, far easier to capture.
It is even more fun to see how Memo reacts to unseen environments. We deploy it to 6 unseen Airbnbs and task the robot with fine-grained tasks such as picking up utensils from the plate. Because we train on data from over 500 homes, the new home is instantly familiar to Memo.
If you want to learn more about ACT-1 and Memo, read our first research blog post: Follow @sundayrobotics for all future updates.
@sundayrobotics Congrats Tony and team!
@sundayrobotics Big flex from Sunday Robotics, hype's real, now show the messy real-world demos
@sundayrobotics Zero-shot generalization to unseen homes!
@sundayrobotics 2000: Honda Asimo iPhone moment (programmed tasks). 2025: @sundayrobotics iPhone moment (full useful autonomous tasks). So sharp job @tonyzzhao ! 2050 will be wild.
@sundayrobotics 2026 would be wild.
@sundayrobotics 💯 20 successful tries without breaking glasses is amazing! Do you use harmonic drives from Nabtesco?
@sundayrobotics Congrats @tonyzzhao @chichengcc and team! Would love to hear the internal conversation of ACT vs diffusion policy :)
@sundayrobotics Cool work @tonyzzhao !
@sundayrobotics Thank you Julian!
@sundayrobotics Congrats!!! Training without data is so exciting. I have a small question: why do some videos look so simulation-like? Is it a specific lighting setup?
@sundayrobotics Congrats Tony on such elegant setup! I know it’s the best one two days ago but I’m still stunned by this video.
@sundayrobotics Building from the ground up offers us so much flexibility: from hardware design, data engine, to model architecture and training objectives. It's been an amazing journey to be learning from experts in my team @sundayrobotics - such a proud moment!
@sundayrobotics I am really excited about what you are building here. And I hope I could someday join you guys.
@sundayrobotics That means a lot to us - Thank you Vector!
@sundayrobotics Congratulations on the awesome release @tonyzzhao!! 🎉 Memo is so cute!!
@sundayrobotics Thank you Kamyar!!
@sundayrobotics LFG!! 🚀🚀 so stoked to be able to share the progress, this is only the beginning here @sundayrobotics 🦾
@sundayrobotics Your robot is cute as hell
@sundayrobotics looks amazing, congrats Tony!
@sundayrobotics Thank you Ayzaan! Still remember the night we shoot this video together. It's been a while!
@sundayrobotics that night was so fun, and so much progress on robotics since then!
@sundayrobotics Yes hard to believe!
@sundayrobotics wow incredible release tony, congrats!
@sundayrobotics Thank you Tyler!!
@sundayrobotics This is so cool Tony, congrats on the launch!
@sundayrobotics Thank you Kent!
@sundayrobotics Amazing work @tonyzzhao !! Excited to see more!
@sundayrobotics Thank you Samuel!
@sundayrobotics Super cool congrats Tony 🚀
@sundayrobotics Thank you Arnav! It's been a while and still missed the days at ML@B!
@sundayrobotics Congrats Tony!! 🚀
@sundayrobotics Thank you Danijar!
@sundayrobotics Dexterity looks impressive here - congrats @tonyzzhao & team!! 🥂
@sundayrobotics Cool, you've come quite far since ALOHA, which was the last thing of yours I saw!
@sundayrobotics We've come a long way as a field!
@sundayrobotics congrats!! @tonyzzhao @chichengcc the robot looks very nice 🖥️
@sundayrobotics Excellent design. Looking forward to seeing this robot in homes around the world!
@sundayrobotics Thank you Rika!
@sundayrobotics Congrats on the release!
@sundayrobotics Thank you Anikait!
@sundayrobotics Super impressive work, congrats @tonyzzhao and @chichengcc!!
@sundayrobotics @chichengcc Thank you Gokul!
@sundayrobotics That sock folding at 3:00 minutes is mind-blowing 🤯 he folded with ease ... Congratulations 👏
@sundayrobotics Great work - also love how the robot looks!
@sundayrobotics Thank you Cyril! Means a lot!
@sundayrobotics Cool, congrats, @tonyzzhao!
@sundayrobotics Thank you Huaxiu!
@sundayrobotics Congrats!!!
@sundayrobotics Thank you Adam!
@sundayrobotics it looks so good!!
@sundayrobotics Congrats Tony, very impressive!
@sundayrobotics Thanks Carlo!
@sundayrobotics So impressive! Congrats!!
@sundayrobotics Thank you Kimin!
@sundayrobotics Congrats on the launch - for the nerds, this video was better than the launch video :)
@sundayrobotics I love @alpercanbe at the end
@sundayrobotics woah 🤯
@sundayrobotics Go easy data collection interfaces (and cute robots)! Congratulations :)
