- Google’s new Gemini Robotics On -Wevice AI model lets robots run completely offline
- The model can learn new tasks from only 50 to 100 examples
- It adapts to different robot types, such as humanoids or industrial arms, and can be used in rural and hospitals
For years, we have been promised robot butlers who are capable of folding your laundry, chopping your onions and essaye a witty bon mot as those in our favorite period dramas. One thing that these promise never mentions is that accidentally removing your router may close the mechanical Jeeves. Google claims its latest Gemini AI model solves this problem.
Google Deepmind has revealed its new Gemini robotics on the device’s AI model as a way of keeping robots safe against abandoned power lines and rural work. While not as powerful as the standard cloud-based gemini models, its independence means it can be much more reliable and useful.
The breakthrough is that AI, a VLA (vision, language, action) model, can look around, understand what it sees, interpret natural language instructions and then act on them without having to look up words or tasks online. When testing, robots with the model have installed completed tasks on unknown objects and in new environments without Google.
It may not look like a huge deal, but the world is full of places with limited internet or no access at all. Robots working in rural hospitals, disaster zones and underground tunnels cannot afford to disappear. Now not only is the model fast, but Google claims it has an amazing ability to learn and adapt. The developers claim that they can learn AI new tricks with as few as 50 demonstrations, which are practically immediate compared to some of the programs currently used for robotic training.
Offline Robot AI
This ability to learn and adapt is also evident in the robot’s flexible physical design. The model was first designed to run Google’s own Aloha devices with double arm, but has since proven to be able to work when installed in far more complex machines, such as Apollo Humanoid Robot from Apptronik.
The idea of machines that learn quickly and act independently raises some red flags. But Google insists it is cautious. The model comes with built -in protective measures, both in its physical design and in the tasks it will perform.
You can’t run out and buy a robot with this model installed yet, but a future involving a robot with this model or one of its descendants is easy to imagine. Let’s say you buy a robot assistant of five years. You want it to do normal things: Fold towels, preparation meals, prevent your toddler from launching the Lego brick down the stairs. But your other child would see how the box with the flashing lights worked, and suddenly these lights stopped blinking. Fortunately, the model installed in your robot can still see and understand what these LEGO bricks are and that you ask it to pick them up and put them back in their bucket.
That’s the real promise of Gemini Robotics on -Wevice. It’s not just about bringing AI into the physical world. It’s about making it stick around when the lights flicker. Your future robot butler will not be a cloudy responsibility. The robots come and they are really wireless. Hopefully it’s still a good thing.
