Control a robot in simulation with the Gemini Robotics Embodied Reasoning model 🤖 Direct tasks like object localization and placement by simply prompting a robotic arm to pick up and place objects.

Get started with the open source physics simulator demo in @googleaistudio:

Build your own “Pick and Place” robotics simulator in your browser with @GoogleAIStudio using this step-by-step guide:

I’m in a unique spot: I teach AI and Industrial Maintenance at Nash Community College, and I’m also a taking my second Robotics class. Seeing Gemini bridge the gap between the code I teach and the physical servos I'm learning to control is mind-blowing. The convergence is here. 🤖🎓

It seems most robotics is mostly about mimicking humans. Personally, I would like to see the kind of robots that produce all the food you need in your back garden. All supplies needed for a comfy life, local to your street.

Tried to create a stack of shirts :(

The gap between text prompts and physical action is closing so fast! 🤯 It's wild to see embodied reasoning handling localization this smoothly. The future of robotics is looking very accessible.

Finally, a robot that can pick things up. My roommate still can't do that after 3 years.

This is huge 🤯 Prompting a robotic arm to localize, pick, and place objects using embodied reasoning is a big step toward truly general-purpose robots 🤖

Claude + webcam... not as advanced, but who knows where it ends up

¿Cómo imaginas la aplicación de esta tecnología en logística internacional? La precisión robótica podría revolucionar cadenas de suministro.

Chefs 🧑‍🍳

Booked will try it out

I LOVE WHEN ALL TEAMS ARE FIRING ON ALL CYLINDERS!!!

Supreme- now make plugins for most prominent industrial robots so we can handle IK, paths, targets in sim real time 🤩🤩

Gemini’s bridging the gap between text prompts and physical actions faster than anyone expected.

This is a big deal if it generalizes beyond curated sim demos. The key bottlenecks I'd love to see numbers on: sample efficiency, robustness to sensor noise, and sim2real transfer. Any public eval suite or benchmark you recommend?

robotics always is pick and place but hope we can get diff drives and legged.

The key thing happening here isn't "prompting a robot arm" — it's that the grounding between language and physical space is working in simulation. Getting sim-to-real transfer to actually work is the hard problem. This demo buys nothing until it moves outside the controlled physics sandbox.

What simulation is that?

Gemini Robotics making simulated control feel real

This is so cool I had built something similar but its no where compared to this

🤖🤖🤖

I am asking myself, which software are you using to simulate the robot?

What's coming with AI + Robotics is just magic

@Google You have to pay me if you want me to do work for you

愿你的所有坚持，都能换来累累硕果👍

@Google @shawmakesmagic

great now i can get rejected by robots in real time instead of just email

ngl this is huge for automation systems... feedback loops in robotics could optimize execution speed ⚡

Help here please

@Google Yo

interesting to see how the gemini robotics embodied reasoning model can be applied to real world scenarios like the helium mobile mart store in new york city where robots could potentially be used to manage inventory and customer service

This is definitely worth the time and effort. Right? ..... Right?

@Google Hey #Google

sounds like the first physical AGI we will get will be an "AI ARM" 😅 ( oor maybe the Terminator Hand ? )

prompting a robot to move objects is wild ngl feels like software finally touching the physical world 🤖