Google DeepMind introduces Gemini Robotics models for physical tasks

Google DeepMind announced Gemini Robotics, a vision-language-action model that outputs robot actions, and Gemini Robotics-ER, a model focused on spatial and embodied reasoning. DeepMind highlighted three goals: adapting to new situations, responding to spoken instructions and changes around the robot, and fine manipulation such as folding origami. It named Apptronik as a partner building humanoids on the technology and listed trusted testers including Boston Dynamics, Agility Robotics and Enchanted Tools. DeepMind also released the ASIMOV dataset for evaluating the safety of robot actions, an important step for robots meant to work around people.

Source: Google DeepMind