A team of engineers from the University of California has designed and created a special robot that has the capacity of learning how to do tasks like humans do: by trial and error.
BRETT the robot represents an important step forward in the field of artificial intelligence.
The scientists designed the robot to work on algorithms that allow it to accumulate its knowledge slowly, in time, just like humans do.
Unlike other robots, which are pre-programmed to do perform things from the moment they are created, BRETT the robot has the capacity of “thinking” on its own.
According to its creators, BRETT the robot has neural nets that were inspired by the human brain circuitry.
Pieter Abbeel, one of the scientists involved in designing and creating BRETT, explained how the robot works, saying that when it needs to learn to do a new task, the robot doesn’t need to be reprogrammed.
The scientists used a software that encodes how the robot can learn new things; the same software allows BRETT the robot to learn different tasks given to it by the engineers.
BRETT, which stands for “Berkeley Robot for the Elimination of Tedious Tasks”, was used in several experiments in which the scientists had it perform a series of motor tasks.
One of the assignments the robot had to do was to put blocks through openings that match and to stack Lego blocks accordingly.
As the robot got closer to the right way of performing the tasks, the scientists “rewarded” it by giving it points allocated by the special algorithm.
Abbeel said that there is still a long way before the robots equipped with artificial intelligence can learn to do things on their own, like cleaning the house or sorting the laundry.
But BRETT the robot indicates that these types of deep learning techniques can prove to be very transforming for the robots, as they enable them to learn how to perform complex tasks completely from scratch.
Household robots need to be able to adapt to the changing conditions of their environment. One way of helping the robots achieve this is by programming them with a wide range of possible scenarios and also giving them instructions on how to do the tasks in each of the cases.
Sergey Levine, one of the researchers involved in the project said:
“For all our versatility, humans are not born with a repertoire of behaviors that can be deployed like a Swiss army knife, and we do not need to be programmed. Instead, we learn new skills over the course of our life from experience and from other humans.”
Image Source: engadget