Google DeepMind has developed a new Gemini Robotics model based on Gemini 2.0, which it says will "lay the foundation for a new generation of helpful robots," Engadget reports.
The model is designed to simplify direct control of robots. The company believes that artificial intelligence for robots should have three key properties: versatility, interactivity, and agility.
Versatility refers to the robot's flexibility, allowing it to adapt to new situations, including those not considered during its training. Interactivity refers to the robot's ability to respond to people and the environment, and dexterity encompasses the ability to perform precise movements related to fine motor skills.
For example, DeepMind’s ALOHA 2 robot, thanks to Gemini Robotics, can fold origami and seal Ziploc bags. This two-armed robot can also understand all instructions given to it in natural, everyday language.
DeepMind also unveiled a model called Gemini Robotics-ER (or embodied intelligence). The company says the model will allow roboticists to run their own programs using Gemini’s advanced reasoning capabilities. DeepMind is giving access to the system to “trusted testers,” including Google subsidiary Boston Dynamics.