Researchers tested first on a group of children and then on iCub, an AI machine: both use the process of mutual exclusion
Sometimes artificial intelligence can be used to understand how the human brain works. According to an experiment, children during the first years of life would apply the same associative mechanism used by robots to learn new words.
This is what emerges from a study conducted on a group of children of two and a half years who were asked to identify a toy among those proposed, which included an object unfamiliar to children. The aim of this test is very simple: to lead the children to learn the name of the unfamiliar toy using the technique of mutual exclusion. According to this psychological (but mainly computer) elimination procedure, children would use the information contained in their memory to label a new object. A bit like AI robots based on deep learning systems.
The experiment carried out with iCub
The children were given three toys: one brown, one yellow and one orange. With regard to the first two, the group most likely already knew which type of animal corresponded to each color. Next, the researchers asked the children to identify the giraffe. The children, applying the mutual exclusion technique, were able to figure out that the giraffe was the orange object.
After performing this test, the study team performed the same experiment on iCub, a child-sized AI robot. The AI system can recognize up to 12 objects. Surprisingly, the machine was also able to detect the unknown toy, applying the same elimination mechanism used by the human group.
This proves one thing, according to the authors of the experiment: in the first years of life, the human brain applies the same simple associative process as robots to learn.