Language-Learning Robot

French scientists have developed a simplified artificial brain that enables robots to learn, understand, and even anticipate what’s being said to them. The model may someday be used to better understand the mechanisms responsible for the linguistic breakdowns that occur in certain diseases, such as Parkinson’s.

The “artificial neuronal network” basically mimics the structure of the human brain and the mechanisms it employs for learning. Through a principle known as recurrent construction, in which connections form in recurring loops, the system continually revises its understanding to recognize sentences even when the grammatical structure changes. It’s also capable of linking sentences and predicting the end of a sentence before it’s provided.

This research has important implications for the field of robotics. Instead of needing to program all of the knowledge that a robot will need, it’s now possible for some of that information to be acquired through a learning process.