Google DeepMind has unveiled two innovative AI models in the Gemini Robotics family: Gemini Robotics-ER 1.5 and Gemini Robotics 1.5. These models are designed to enhance the capabilities of general-purpose robots, improving reasoning, vision, and action in dynamic real-world scenarios. The two-model architecture separates planning from execution, addressing the shortcomings of single-system AI.
Gemini Robotics-ER 1.5 acts as the planner, utilizing advanced vision-language processing to develop multi-step plans and access external resources like Google Search for informed decision-making. Conversely, Gemini Robotics 1.5 executes these plans, converting natural language instructions and visual data into precise motor commands and explaining its choices in everyday language.
This dual system allows robots to effectively manage complex tasks, such as sorting waste according to local recycling guidelines. Available through the Gemini API in Google AI Studio, these adaptable models signify a leap toward integrating generative AI into robotics, offering a natural language interface for enhanced performance.
Source link
