Murtaza Dalal
@mihdalal
Followers
8K
Following
498
Media
37
Statuses
304
AI @ Tesla Optimus | Robotics/AI PhD @CMU_Robotics, EECS @UCBerkeley | Opinions entirely and unequivocally my own
San Jose, CA
Joined February 2020
I want to make clear how crazy impressive this result is. We can now do bi-manual, dexterous manipulation across a wide range of tasks with barely any data on these skills coming from teleoperation. As we know, teleop does not scale! But turns out human video does!. This means.
164
353
4K
Incredibly simple method to induce novel task solving during inference time: warm start denoising with human priors!.
Presenting DemoDiffusion: An extremely simple approach enabling a pre-trained 'generalist' diffusion policy to follow a human-demonstration for a novel task during inference. One-shot human imitation *without* requiring any paired human-robot data or online RL 🙂. 1/n
2
0
9
Taking “teleop” to the next level 🧠 🤖. Jk but like very cool result didnt think it was possible to do this.
3
4
78
As someone who has often just done this based intuition, this would be super useful in general ngl. Would love to see a slightly more formal treatment.
I wish there was a good venue to write/present about RL “tricks”. how PD gains affect action scale, how to tune reward functions, actor STD, etc. there’s good intuition for all of it, grounded both in learning theory and robot dynamics, but I don’t often see good explanations.
2
2
17
Hopefully one day we will have FSD deliver Optimus to customer doorsteps too :P. Or a legion in a Tesla Semi, that would be cool.
The first fully autonomous delivery of a Tesla Model Y from factory to a customer home across town, including highways, was just completed a day ahead of schedule!!. Congratulations to the @Tesla_AI teams, both software & AI chip design!.
12
11
236
Active perception is a critical skill for robotic manipulation!.
Your bimanual manipulators might need a Robot Neck 🤖🦒. Introducing Vision in Action: Learning Active Perception from Human Demonstrations. ViA learns task-specific, active perceptual strategies—such as searching, tracking, and focusing—directly from human demos, enabling robust
1
2
21
RT @Jiahui_Yang6709: Yes!! Thrilled to present Neural MP at IROS🥳. We’ve also been pushing this line of work even further! Stay tuned, big….
0
1
0
RT @chris_j_paxton: Cool work on neural motion planning. much more challenging problem in a lot of ways than vision-language models - becau….
0
2
0
Incredibly excited to share that Neural MP got accepted to IROS as an Oral presentation!! Huge congrats to the whole team (@Jiahui_Yang6709, @mendonca_rl , youssef khaky, @rsalakhu, @pathak2206), but especially @Jiahui_Yang6709 for making this happen after I graduated! This now.
Can a single neural network policy generalize over poses, objects, obstacles, backgrounds, scene arrangements, in-hand objects, and start/goal states?. Introducing Neural MP: A generalist policy for solving motion planning tasks in the real world 🤖.1/N
3
9
90
Massive congrats to our neighbors in the office (aka the autopilot team) for making this happen! As an avid FSD user I'm especially excited, this is just getting us one step closer to the future where we don't have to pay attention to the road at all and just let the car drive!.
BREAKING: First ever Tesla Model Y robotaxi with no-one in the drivers seat spotted testing on public roads in Austin, Texas!. Tesla's new "Robotaxi" wordmark/logo is on the side of the vehicle.
11
37
787
Really cool work from @g_k_swamy and co! Sample efficiency gains by leveraging search in world models is a cool way to enable test time reasoning for robotics!.
Say ahoy to 𝚂𝙰𝙸𝙻𝙾𝚁⛵: a new paradigm of *learning to search* from demonstrations, enabling test-time reasoning about how to recover from mistakes w/o any additional human feedback! 𝚂𝙰𝙸𝙻𝙾𝚁 ⛵ out-performs Diffusion Policies trained via behavioral cloning on 5-10x data!
0
3
28
The place I least expected to see this 😂. But hey looks like e-trade is getting with the program!
4
0
23
RT @julianibarz: Wishing my colleague and friend, @milankovac, all the best in his next chapter; mais avant, des vacances bien meritees! ;)….
0
4
0
I’ve only been at Tesla for a few months, but I’ve never been more impressed with a leader. Not only are you incredibly knowledgeable about pretty much every technical part of Optimus - from hardware mechanical and electrical to even AI, (that’s honestly kinda insane if you think.
Over the past 9+ years, I’ve had the immense privilege to work with some of the most brilliant minds in AI & engineering. I’ve built friendships that will last a lifetime. This week, I’ve had to make the most difficult decision of my life and will be moving out of my position.
11
51
942
Yeah this was the challenge I was trying to address during a large portion of my PhD as well. I had a similar intuition/finding that horizon is the problem - with a different set of solutions. What worked well in the end: Locally robust pre-trained skills + motion planning + VLM.
The fact that current RL methods struggle at long-horizon tasks is something that I think most people understand intuitively; this is a really cool paper and bit of analysis explaining what's going on, and how to work around it. This remains to me an important issue that needs.
3
1
49
Great work by @ZhaoMandi and an excellent paper title 😛. Looking forward to a real robot follow up sometime soon I hope.
How to learn dexterous manipulation for any robot hand from a single human demonstration?. Check out DexMachina, our new RL algorithm that learns long-horizon, bimanual dexterous policies for a variety of dexterous hands, articulated objects, and complex motions.
0
3
23
This is why we do what we do :).
"Robotics is going to be the biggest industry in the history of the planet. There's going to be. hundreds of billions of robots of all shapes & sizes running around doing all kinds of things". Don't take my word for it - hear it straight from @pmarca
2
4
127
Yeah its the grit + not really being afraid of hard problems anymore.
the most satisfying takeaway from a phd is that you can solve problems far beyond your capabilities if you're willing to throw yourself at them again and again, over a long period, while staying open to new ideas. it's not really about being smart. just curious & persistent.
1
0
12