Google has created a “robot constitution” to guarantee human safety
Google DeepMind team presented three developments that will help robots make faster, better, and safer decisions, reports The Verge.
One of them is the AutoRT data collection system. It can use a visual language model (VLM) and a large language model (LLM). Working together, they help robots understand their environment, adapt to unfamiliar conditions, and make appropriate decisions.
Interestingly, this development is based on the Robot Constitution. It is meant to ensure that the user assistant does good, not harm. This is actually a set of “safety-oriented tips”.
The authors of the Robot Constitution were inspired by Isaac Asimov’s Three Laws of Robotics. They are as follows:
- A robot cannot harm a person or, through its inaction, allow a person to be harmed;
- A robot must obey human orders when these orders do not contradict the First Law;
- A robot must take care of its safety as long as it does not contradict the First and Second Laws.
Within seven months, Google deployed a fleet of 53 AutoRT robots in four different office buildings and conducted more than 77 thousand tests.
Some robots were controlled remotely by human operators, while others worked either scripted or fully autonomously using Google’s Robotic Transformer (RT-2) artificial intelligence learning model.
Another DeepMind development is the SARA-RT neural network architecture. It was developed to make the existing RT-2 transformer robot more accurate and faster. The company also announced RT-Trajectory, which adds 2D shapes to help robots better perform specific physical tasks.