CALIFORNIA (Web Desk) - Tech giant Google has formulated a 'Robot Constitution' to avoid mishaps caused by robots.
Google is hopeful that its DeepMind Robotics Division will create such a helper robot that can follow commands like cleaning the house or cooking a meal. But a seemingly simple request can turn out to be dangerous with the robot's understanding.
This implies that a robot might misinterpret the commands. It might be unaware that the house should not be cleaned in a way that destructs home.
Google has introduced a set of new systems, and by following them, the company will be able to create robots that will assist humans without potential harm.
The company believes the vision behind these new systems is to make robots that have the qualities of swift decision-making, better understanding of the environment around them, and effective navigation.
The breakthrough includes a new system called ‘Auto RT’ that uses artificial intelligence to understand human intent using a large language model (LLM).
The system takes data from cameras mounted on the robot and feeds it to a ‘Visual Language Model’ (VLM), which can understand the environment and the objects in it and describe it in words.
This system sent this data to LLM, which understands these words, lists the possible tasks, and then decides which tasks should be performed.
However, Google also made it clear that in order for these robots to become part of their daily lives, people will need to ensure that they behave consciously and safely. Therefore, Google has fed the ‘Robot Constitution’ in the Auto RT’s decision-making part, LLM.
According to Google, these sets are safety-based instructions that robots follow when choosing tasks.