Google's Leap into Robotics with Gemini-Based AI Models

In a groundbreaking announcement on Wednesday, Google introduced two new artificial intelligence models, both based on the advanced Gemini 2.0 technology, specifically designed to revolutionize the field of robotics.

Introducing Gemini Robotics and Gemini Robotics-ER

The first of these models, Gemini Robotics, is an advanced vision-language-action (VLA) model. It represents a new output modality for directly controlling robots, thanks to its unique addition of physical actions. The second model, Gemini Robotics-ER, comes equipped with advanced spatial understanding capabilities. This allows roboticists to leverage Gemini's embodied reasoning (ER) abilities to execute their own programs, opening up new possibilities in robotics.

Collaboration with Apptronik for Next-Gen Humanoid Robots

In an exciting development, Google also revealed its partnership with Apptronik. Together, they aim to build the next generation of humanoid robots utilizing the Gemini 2.0 technology. While the Gemini Robotics model was primarily trained on the bi-arm robotic platform ALOHA 2, Google has stated its potential for specialization in more complex embodiments, such as Apptronik's humanoid Apollo robot.