Pedro Milcent

What if Physical AI could fully leverage human videos for model training? 💪➡️🦾 Human demos like the one below hold massive potential, but to unlock that power, we need to extract the relevant information they contain.

Great event today at the Robotics Symposium (BARS) at @Stanford . Amazing to see how much has evolved just this year.

That’s what our Motion2Text API is for: turning human demos into fine-grained language descriptions that guide robotic models and make human data truly useful. Interested in trying out? Get in touch: forms.gle/sSHTPBDVkBtnHh…

What if Physical AI could fully leverage human videos for model training? 💪➡️🦾 Human demos like the one below hold massive potential, but to unlock that power, we need to extract the relevant information they contain.

@aesposito__ @IlirAliu_ @DeplaceAI Also see a lot of value in Motion2Text for 3rd person videos, something we will focus more on later. Today, we're working on language to better leverage ego human demos for contact points, trajectory strategies, etc.

@IlirAliu_ @DeplaceAI motion to language to motion seems unnecessary for first-person videos. it might make more sense for third-person, at least until someone cracks video-to-simulation

You will soon be able to teach robots what human are doing… using natural language. I spoke to one of the founders a couple of months ago on my podcast: An API that takes raw human videos and returns detailed language annotations describing the motion. @DeplaceAI is building something wild: Not just basic labels, it captures: ✅ motion semantics ✅ relative positions ✅ cause and effect ✅ task outcomes ✅ and more It’s built on top of research in point tracking, segmentation, and learning from demos; all to make it easier to train robots and embodied agents without manual labeling. They’re offering early access and sharing some of the datasets they collected via their global network of video collectors. Thanks for sharing, @MilcentPedro ! 🔗 forms.gle/sSHTPBDVkBtnHh… One of the most interesting motion-to-language interfaces I’ve seen. If you’re working in robotics, vision, or LfD, it’s worth checking out.

@IlirAliu_ @DeplaceAI There’s a lot of info in human demos: object affordances, trajectory strategies, contact points, etc We just have to make it useful for robotics

✋➡️🤖 human intelligence to robotic intelligence through language

What if you could send human videos and get back detailed language descriptions of the motion to guide your Physical AI model? Happy to share more about the Motion2Text API we are building at @DeplaceAI! We are working on natural language descriptions that capture: 📙 Motion semantics 📏 Relative positions ⚙️ Cause and effect 💡 and more Leveraging state-of-the-art research and models in point tracking, semantic generalization, object segmentation, and learning from human demos ✋➡️🤖 This helps robots understand human movements and close the gap between humans & robots to fully leverage human videos to train Physical AI! 👉 If you want early access to the API or to use the datasets we gathered through our network of collectors, let us know, we would love your feedback: lnkd.in/eM5i4Zcc

@aaronistan Impeccable taste 🤌

@MilcentPedro Thank you! Did you like the Parisian architecture in the video

Introducing Lume, the robotic lamp. The first robot designed to fit naturally into your home and help with chores, starting with laundry folding. If you’re looking for help and want to avoid the privacy and safety concerns of humanoids in your home, pre-order now.

Interesting research from @ylecun and the @metaai + @Mila_Quebec teams on a task-agnostic, action-conditioned world model 🌐 The paper has valuable insights into training strategies and model design. Check out V-JEPA 2 here: arxiv.org/pdf/2506.09985

Awesome few days in Paris, @ycombinator event at Sorbonne had great tips for early-stage founders, and the @RaiseSummit was all about the future of AI (Physical AI too!). More tomorrow 🦾 🤖

@NEURARobotics Great to see more humanoid makers in Europe 🇪🇺

What a reveal! 🦾 The new generation of 4NE1 is finally here. Ready to bring safe, intelligent automation to the real world. Thanks to everyone who was there live for this big moment at automatica 2025. More impressions coming soon. Stay tuned!

@Stone_Tao @BoAi0110 Will check it out! Thanks a lot 💪💪

@MilcentPedro @BoAi0110 embodiment-scaling-laws.github.io see the website, there’s a presentation there. @BoAi0110 will post more info later as well

Rest of my lab has various presentations at RSS2025, please check out their awesome work! June 25 EEB 248 @ 10am - Hardware Optimization for In-Hand Rotation by K. Fay June 25 EEB 248 @ 3pm - Towards Embodiment Scaling Laws in Robot Locomotion by @BoAi0110 June 25 OHE 122 @ 11am - ImVR: Immersive VR Teleoperation System for General Purpose by Yulin Liu

Great paper by @mimicrobotics, showcasing the 16 DoF Faive hand in action and highlighting the importance of #real, diverse, and curated #data for Physical AI models that are performant, generalizable, and capable of self-correction 📈✋ 📄 Check out the paper by @elvisnavah and team here: arxiv.org/abs/2506.11916

@xiao_ted Would love to watch the session if it's recorded, big momentum in Physical AI data 🦾

Excited to attend #RSS2025 in LA this coming week! Lots of progress in the field on embracing data-driven robotics, I’m expecting a very different vibe this year. I’ll be giving a talk on Gemini Robotics and debating at the SemRob and RoboEval Workshops, hope to see you there!

@peide_huang @SnehalJauhri @GeorgiaChal @GalassoFab10 @danfei_xu @YuXiang_IRVL @ryan_hoque Well-deserved 👏 Many insights from the paper were very helpful for the human demo collections we’re doing at @DeplaceAI

Very happy that EgoDex received Best Paper Awards of 1st EgoAct workshop at #RSS2025! Huge thanks to the organizing committee @SnehalJauhri @GeorgiaChal @GalassoFab10 @danfei_xu @YuXiang_IRVL for putting out this forward-looking workshop. Also kudos to my colleagues @ryan_hoque David Yoon, Mouli Sivapurapu, @jian_zhang_ !

@RoboticsSciSys Amazing to see so many of the top minds in robotics at the USC campus! 🦾🦾

🌟 Day 2 at #RSS2025! 🌟 We just had a super inspiring Early Career Spotlight Talk by Dorsa Sadigh: "From Dirt to Data: How Gardening Taught me about Generalist Robot Policies"! 🤖💡 Stay tuned for more excitements today!

@SnehalJauhri Thanks for sharing the recording! Both papers are extremely relevant to scale data collection for robotics 👏👏

Thank you to all the speakers & attendees for making the EgoAct workshop a great success! Congratulations to the winners of the Best Paper Awards: EgoDex & DexWild! The full recording is available at: youtu.be/64yLApbBZ7I Some highlights:

@elvisnavah Awesome results, really shows how crucial data collection decisions are for model performance!

Happy to announce mimic-one: a Scalable Model Recipe for General Purpose Robot Dexterity, the culmination of years of research work in dexterous manipulation with imitation learning.