train neural nets to make robot do useful stuff at @Tesla_Optimus

California, USA
Joined March 2018
exited to share what I’ve been working on with @mendonca_rl and many other team members in the past a few months: teaching Optimus new tasks by watching human videos! One step closer to real world agi!
I’m not just dancing all day, ok
19
12
1
484
28,288
“who is safe?” “claude is SAFE!”
“请Dario看到视频后把我列入Claude白名单,别封我” 歌曲名:“Claude is Safe!” 歌词、视频预处理:GPT-6-Astra 作曲:Suno V6 图像生成:GPT-Image-2.5 视频模型:Seedance2.5 参考视频:Michael Jackson - “Dangerous World Tour”
2
1,368
most dexterous teleop I’ve seen in a while from @SharpaRobotics
sharpaのデモすごいな。VRで手先の位置指令+ハンドは振動でフィードバック。操作者の技術も極まってる。
3
6
45
10,160
humanoid arena 🙃
Is Opus 5.5 better than GPT-6 Astra at Physical Intelligence "literally" ? put GPT-6 Astra and Claude Opus 5.5 in two Unitree G1s in MuJoCo Ring Fence. Every 2 sim seconds: state → LLM → joint targets → act. Goal: force the other out of the ring, and Astra won after 31 second.
1
9
2,083
when you work on something deep and long enough, there are things that extremely obvious to you but extremely not obvious to others
4
4
1
40
2,952
you can look around and then close your eyes for 3 seconds; you’d be surprised how many things you can do within this 3 seconds of blind
1
1
7
837
first UMI + humanoid demo I’ve seen. looks promising 🙂
Introducing OM-1, our first robot foundation model, zero-shot generalizing to any robot: table-top arms, industrial arms and humanoids. - learned directly from human manipulation data - no teleop/robot data - close to human-level dexterity and efficiency - multi-robot collab
4
2
26
4,302
damn, just flat tire the 3rd time this year by myself. I guess I never understand the model Y collision geometry better than FSD 🥲
9
948
I’ll eat an egg without peeling the shell, if some one show me a video of GPT-6 teleop a humanoid doing loco-manipulation stuff 🙃
I didn't get what GPT-6 meant for robotics until I actually tried it. GPT-6 Astra just does physical ICL out of the box. we drop a recording of a human doing a novel task into 𝗰𝗼𝗱𝗲𝘅 app. Prompt it to drive a robot arm the same way. It just works on the first pass!
8
46
9,171
reminds me The Animatrix “World Record”
the craziest moment from China’s AI robot race
730
My X account was hacked earlier today. Please don’t click the message asking for vote (see the attached photo). I’m sorry if this brings you any inconvenience 🙏
2
17
1,301
reading is discriminative writing is generative let that sink in
1
6
1,002
is it me or anybody else feel like claude is talking in less plain English recently…?
8
893
“using open model to defend attacks from close model” I bet this is gonna happen in the future for robotics/humanoid as well 😆
We got attacked by secret unreleased proprietary models and defended ourselves with an open model, more precisely the @nvidia quantized version of GLM 5.2 coming from @Zai_org. Banning any open model would hurt first cyber security defenders, startups, small companies, researchers and everyone who's not a frontier lab and need on-prem affordable controlable models to compete and protect themselves. Let's not do that!
2
1,251
this looks to me that, all end effectors are just another tool on top of dexterous hands
Why create robot intelligence for just one hand, when we could have it learn from many? GEN-1, our latest embodied foundation model, now supports a broad range of end effectors from 5-finger hands, to specialized tools, and everything in between.
5
1,181
so exoskeleton in gloves?
The natural final evolution of UMI devices will be the human hand form factor. Why teleoperate when you can just wear the end-effector like a glove.
1
1
13
3,968
There are probably two most important books for Americans: Constitution and Bible. I bought the first one a few years ago, as I’ve never read it since I came to America 10+ years ago. And I want to read it thoroughly other than anecdotal quotes from internet. I was a bit shocked by how straightforward and pragmatic this book is, it reads almost like an instruction/MD file as the founding fathers’ best practices to run a country based on simple principles: separation of powers, protection of individual property, freedom of speech, equal right to pursuit happiness, et al. And it is short, I finished reading it during a flight, kept it in my backpack ever since! And it works, for at least 250 years Happy 4th of July
1
1
18
1,585
just realized claude/codex can’t replace me yet, if they can’t watch screen/videos 🤔
3
673
there are only two predictions matters, predicting states and predicting actions, everything else follows
One of the bitter lessons in computer vision is the false assumption that solving all the intermediate representations—segmentation, 3D meshes, depth, etc.—is needed to make a computer understand the scene. In the end, it is the final policy that takes action that matters.
2
10
3,084
so why is he walking like this 🙃
9
1
32
4,608