Jun Zeng

Ph.D. in Robotics @UCBerkeley, X2014 @Polytechnique. Love mathematics, robotics and programming.

[Jeff Dean @JeffDean]

Announcing Discovery Loop!

I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the world’s most used products, infrastructure and AI models, and we’re excited to turn our attention to this ambitious endeavor.

♾

Learn more at: https://t.co/Rv3LMdLluK

[Jun Zeng @junzengx14]

@yswhynot @physical_int Congrats!

[Jun Zeng @junzengx14]

@xiaolonw @Meta Wow congrats!!

[Jun Zeng @junzengx14]

@imawasif It actually grabbed my phone on that day!

[Jun Zeng @junzengx14]

Recently found a photo taken by my iPhone 5s many years ago at TechCrunch, the field has evolved so fast for the past few years. 🧐

[Jun Zeng @junzengx14]

@VicGriffith1001 Agree 踏空比亏钱还要难受。。。

[Jun Zeng @junzengx14]

@GhaffariMaani @ToyotaResearch @JustinTzuYuan @SangbaeMIT @CynthiaChienErh @UMRobotics @SangliTeng @SangliTeng You smiles so happily 🙂‍↔️

[Jun Zeng @junzengx14]

@VicGriffith1001 看起来是个motion capture系统

[Jim Fan @DrJimFan]

We RL'ed humanoid robots to Cristiano Ronaldo, LeBron James, and Kobe Byrant! These are neural nets running on real hardware at our GEAR lab. Most robot demos you see online speed videos up. We actually *slow them down* so you can enjoy the fluid motions.

I'm excited to announce "ASAP", a "real2sim2real" model that masters extremely smooth and dynamic motions for humanoid whole body control.

We pretrain the robot in simulation first, but there is a notorious "sim2real" gap: it's very difficult for hand-engineered physics equations to match real world dynamics.

Our fix is simple: just deploy a pretrained policy on real hardware, collect data, and replay the motion in sim. The replay will obviously have many errors, but that gives a rich signal to compensate for the physics discrepancy. Use another neural net to learn the delta. Basically, we "patch up" a traditional physics engine, so that the robot can experience almost the real world at scale in GPUs.

The future is hybrid simulation: combine the power of classical sim engines refined over decades and the uncanny ability of modern NNs to capture a messy world.

[Jun Zeng @junzengx14]

@VicGriffith1001 Asuka是个海学人士观点啊🤔

[Jun Zeng @junzengx14]

@VicGriffith1001 第一次来听这种 过于抽象笑不活了🤣

[Jun Zeng @junzengx14]

@ZhongyuLi4 Finally you did this! We chatted about this long while ago haha!

[Jun Zeng @junzengx14]

Amazing work from @qiayuanliao!!!

[Zhongyu Li @ZhongyuLi4]

Introducing the Berkeley Humanoid which was born in our lab! Built from the ground up for "modern" legged control (RL😆), it has a very small sim2real gap. What you see in sim is what it can do in real! It is open for research collaborations - DM us if you want to try your algo!

[Jun Zeng @junzengx14]

@ZhongyuLi4 The one for illegal immigrant workers or contractors is usually lower than this.

[Jun Zeng @junzengx14]

@LarsLindemann2 Congrats! USCIS is always too bureaucratic lol.

[Tianyi Zhang @tianyiz2022]

[DarkGS] Building 3D Gaussians with a torch🔦 in the dark🌃
paper: https://t.co/okH7JeCQft
code: https://t.co/zMOXPPgvyj

[Jun Zeng @junzengx14]

@lucacarlone1 Congrats!! 🎉

[Jun Zeng @junzengx14]

@ZhongyuLi4 I would say since warehouse has been initially designed for large machinery / electric devices instead of human, thus maybe not for bipedal ones also in some common senses 🤔

[Ryohei Sasaki@engineer @rsasaki0109]

DRIFT: Dead Reckoning In Field Time
https://t.co/Eqa3ZodzP2
> Dead Reckoning In Field Time (DRIFT) is an open-source C++ software library designed to provide accurate and high-frequency proprioceptive state estimation for a variety of mobile robot architectures.