Google Deepmind introduces Gemini Robotics 1.5 - Watch Robots Plan, Analyze and Act

Google Deepmind introduced two Gemini Robotics AI models, ER 1.5 and 1.5, to improve the general purposes. ER 1.5 Plan tasks using reasoning and tools, while 1.5 performs from visual inputs and instructions, enabling complicated multi-step tasks with natural language control. Google Deepmind introduced two new artificial intelligence (AI) models in its twin-Robotics family, which aims to improve the capabilities of general robots. (Google Deepmind) Google Deepmind has introduced two new artificial intelligence (AI) models in its Gemini-Robot family, which aim to promote the capabilities of general-purposes robots. The models, called Gemini Robotics-ER 1.5 and Gemini Robotics 1.5, are designed to work together to improve reasoning, vision and actions in the actual environments. Two-model system for planning and execution according to a blog post of Deepmind, twin robotics-er 1.5, serves as the planner or orchestrator, while Gemini Robotics 1.5 is responsible for performing tasks based on natural language instructions. The two-model system is intended to address restrictions seen in earlier AI models, where a single system has performed both planned and planned, which often leads to errors or delays in the execution. Gemini Robotics-ER 1.5: The planner the ER 1.5 model functions as a vision language model (VLM) that can have advanced reasoning and tool integration. This can create multi-step plans for a given task and it is reported that it performs strongly on spatial understanding. The model also has access to external instruments, such as Google Search, to gather information for decision -making in physical environments. Gemini Robotics 1.5: Task performance Once a plan is formulated, twin robotics 1.5, a vision-language action (custard) model, translates instructions and visual input into motor assignments, enabling the robot to perform the task. The model assesses the most effective path to complete an action and performs it, and also provides explanations of decision making in natural language. The handling of complex multi-step tasks The system is designed to handle robots in a seamless process complex, multi-step commands. For example, a robot can sort items in compost, recycling and trash can after consulting the local recycling guidelines online, analyzing the objects, planning the sorting process and then performing the actions. Deepmind mentions that the AI ​​models are adaptable to robots of different shapes and sizes due to their spatial awareness and flexible design. At present, the orchestration model, Gemini Robotics-ER 1.5, is accessible to developers via the Gemini API in Google AI Studio, while the custard model is limited to selected partners. This development is a step into the integration of generative AI into robotics, and replaces traditional interfaces with natural language -driven control, and also tries to separate the planning of execution to reduce mistakes.