A new approach that enables machines to learn how systems (natural and artificial) work and behave by observation has been developed by researchers at the University of Sheffield in the U.K.
Called "Turing Learning," the approach is based on and inspired by the Turing test that was developed by British computer scientist and mathematician Alan Turing in the 1950's.
In a Turing test, a human is required to watch over and interrogate two players - one machine and one human - to tell them apart. If the interrogator is unable to tell them apart, the machine passes the test and has reached the level of human intelligence.
The Turing Learning, on the other hand, employs a computer program that is expected to learn by itself how to interrogate two groups of swarm robots and to figure out how they work.
One of the advantages of Turning Learning, according to Dr. Roderich Gross from the Department of Automatic Control and Systems Engineering at the university, is that machines are no longer told what to look for, and they simply try to understand on their own the reality of how the systems work.
"Imagine you want a robot to paint like Picasso. Conventional machine learning algorithms would rate the robot's paintings for how closely they resembled a Picasso. But someone would have to tell the algorithms what is considered similar to a Picasso to begin with." said Dr. Gross. "Turing Learning does not require such prior knowledge. It would simply reward the robot if it painted something that was considered genuine by the interrogators. Turing Learning would simultaneously learn how to interrogate and how to paint."
With Turing Learning, Dr. Gross also said that computer games could one day have the ability to act as "virtual players," observing and acquiring similar skills of humans. The approach could also be used in many other areas such as in health monitoring and maintenance of vehicles and aricrafts.