Google DeepMind has unveiled two advanced AI models, Gemini Robotics and Gemini Robotics-ER, designed to enhance robots’ abilities to perceive, reason, and act in real-time.

Gemini Robotics leverages the capabilities of the Gemini 2.0 model to enable robots to comprehend and navigate diverse scenarios, even those they haven’t been explicitly trained for. This advancement allows robots to perform intricate tasks such as folding origami figures, organizing workspaces based on verbal instructions, and managing delicate operations like wrapping headphone wires.

Gemini Robotics-ER is an enhanced visual language model that equips robots with sophisticated reasoning abilities, enabling them to tackle complex tasks like efficiently packing a lunchbox. Safety remains a paramount concern; these AI models are trained to evaluate the safety of actions before execution, ensuring responsible operation.

To accelerate the integration of these models into practical applications, Google DeepMind is collaborating with robotics companies such as Apptronik. This partnership aims to develop humanoid robots that are more intelligent, responsive, and capable of adapting to various environments and tasks.

These developments signify a substantial leap toward creating robots that can seamlessly interact with and adapt to real-world environments, performing tasks with a level of dexterity and understanding that closely mirrors human abilities.