Gemini Robotics can teach robots how to perform everyday tasks, even if they have not been trained for that
Share this article
THE Google DeepMind (Big Tech Research Division) is releasing a version of Gemini Robotics to robotic devices. Google AI model serves to ‘teach’ new tasks to robots, even if they have not been specifically trained for it.
The advantage of the system is that it works without internet connection, and can be used in low connectivity training sites or robots that work in strict protocol environments.
Gemini will operate robots
The model GEMINI ROBOTICS It was released in March. The goal is Help robots perform a wide range of real -world tasks, even if they have not been trained for these utilities.
To do this, Google AI uses a Language-Action (VLA) model, which allows you to ‘see’ situations and understand them. In that case, Even if the device has been trained for a very specific activity, it can understand new situations and respond to commands with motor qualities.
The news is that the model can now be applied directly to robotic devices. According to Big Tech, the model is “small and efficient enough to be performed directly on a robot”, allowing us to expand their utilities in everyday life.
AI model will work without internet
One of the advantages is that the model has offline features. According to Google Deepmind robotics chief Carolina Parada, Gemini Robotics has a hybrid approach that allows you to operate both cloud and locally on the device.
To the website The VergeShe revealed that the equipped robot can perform various tasks and adapt to different situations “with only 50 to 100 demonstrations.” It may seem a lot, but it is a small number if we consider a machine that has not been trained for it.
Another advantage is that the model works on several different robotic devices. Google trained its own robot model, Aloha, but also managed to apply the system to machines from other companies, such as Apollo, from AppTonik, and Bi-Bi-Bi-Bi-Bi-Bi-Bi-Bi-Bi-Bi-FR3.
Read more:
What is the gemini for robots for?
- Second stop, the model is considered initial or ideal for poor connectivity applications;
- In this case, the system can be used in places with low connectivity or by companies with strict safety requirements (such as cases where even the lack of internet cannot interrupt the operation);
- As you may have noticed, this version of Gemini is not intended for the common public. It will be available to a group of selected testers;
- Meanwhile, Google is still working to minimize possible safety risks.
Vitoria Lopes Gomez is writer in the digital look