Google DeepMind is committed to improving the decision-making capabilities of robots in complex environments. It has recently released three important developments, the most eye-catching of which is the "Robot Constitution" data collection system. The system combines visual language models and large-scale language models to provide safety tips for robots to ensure that they do not cause harm to humans when performing tasks. This move aims to improve the safety and reliability of robots, further promote the development of robot technology, and make it more widely used in various fields.
Google DeepMind's robotics team has released three new developments aimed at improving robots' decision-making capabilities in complex environments. One of them is a data collection system equipped with a "robot constitution" designed to ensure that robots do not harm humans when performing tasks. The system uses a visual language model and a large language model to provide safety tips for the robot by understanding the environment and task requirements. In field tests, Google deployed 53 robots and adopted measures such as joint force thresholds and physical kill switches to ensure safety. These advances help advance robotics technology, making it more adaptable to a variety of tasks and environments.
These three developments demonstrate Google DeepMind's efforts in robot safety and practicality. In the future, it is expected to see smarter and safer robots used in more scenarios to bring convenience to human life. The field test of 53 robots also provided valuable data support for subsequent research.