It is likely that very soon robots willliving in homes with people, helping the elderly to live independently. But for this they will have to learn how to do all the little work that people could do without thinking. Many modern systems of artificial intelligence are trained to perform certain tasks by analyzing thousands of signed images of a particular action. Although these methods help to solve more and more complex tasks, they still deal only with very specific tasks and require a lot of time and computing power for training.
If the robot will take care of the elderly,problems of such work will vary greatly when compared with typical situations in the learning process. During the day, robots have to do a lot of things, from making tea to changing bed linen during a conversation. These are complex tasks that become more difficult in combination. There are no two identical houses, which means that robots will have to quickly learn and adapt to the environment. And as often happens, if you live with someone else, things tend to migrate. The robot will have to learn how to find them yourself.
One approach is todevelop a lifelong learning robot that can store knowledge based on experience and develop ways to adapt and apply them to new tasks. After you learn how to make a cup of tea, these skills can be applied to coffee.
The human brain learns throughout itslife, constantly adapting to complex and changing conditions and solving a wide range of problems every day. Modeling how people learn can help design robots with which we can interact in a natural way, as if with another person.
Child development simulation for learning robot
The first question to ask when you start modeling people is where to start? Alan Turing, a famous mathematician and pioneer of artificial intelligence, once said:
"Instead of trying to create a programto simulate adult consciousness, why not try to produce a program to simulate a child? If it were so, then after an appropriate course of education, one could get an adult brain. ”
He compared the brain of a child with an empty notebook.a book that can be filled in the process of education and develop an intellectual adult "system". But what should be the age of a child for modeling? What knowledge and skills need to be laid first?
Newborn babies are very limited in thatcan do and how to perceive the world around them. Muscle strength in the child’s neck is not enough to support the head, and he has not learned to control his arms and legs.
Starting from the zero month - such a move can greatlyrestrict the robot. But the child’s physical limitations actually help him focus on solving a small subclass of tasks, for example, he learns to relate his eyes to what he hears and sees. These steps in the initial stages of building a model of the child build up his body before he begins to understand the complexity of the world around him.
Engineers applied similar restrictions to the robot,initially blocking the movement of various joints to mimic the lack of muscle control. They also corrected the images from the camera of the robot, so that it “saw” the world through the eyes of a newborn - with blurring and weak peripherals. Instead of telling the robot how to move, he was allowed to find out on his own. The advantage is that as the calibration changes, or as the limbs are damaged, the robot will be able to adapt to these changes and continue working.
Learning by playing
Studies have shown that through the use oflimitations in the learning process, not only increases the speed with which new knowledge and skills are acquired, but also increases the accuracy of what is being learned.
Giving the robot control of removalrestrictions - giving him control over the joints and improving his eyesight - it is possible to ensure that the robot itself will control its learning speed. Scientists modeled the "child" and the first 10 months of its growth. As the robot learned to relate the movements to the sensory information it received, it acquired the stereotypical behavior observed in babies — such as when children spend long periods of time staring at their hands as they move.
When a robot learns to coordinate itshis own body, the next important milestone he passes - he begins to understand the world around him. The game is an important part of the child’s learning. She helps him explore the environment, test various possibilities and study the results.
At first it may be something simple, likeknock with a spoon on the table or put some object in your mouth. But then it develops into the construction of cubic towers or the placement of objects into suitable openings. All these actions create an experience that will later provide the basis for skills such as finding the right key to open the lock and fine motor skills, to insert the key into the keyhole and then turn it.
In the future, using these methods will give robotstools for learning and adapting to difficult conditions and tasks that people take for granted in everyday life. One day, robots will be able to help older people, but even children in kindergarten will be able to teach them.
Do not forget that you can discuss robots in our chat in Telegram.