iCub robot learns language like baby

A team working with the iCub humanoid robot says it can now understand what is being said to it and even anticipate the end of a sentence.

The Inserm-led team’s ‘simplified artificial brain’ reproduces certain types of so-called recurrent connections observed in the human brain. It can learn, and subsequently understand, new sentences containing a new grammatical structure, can link two sentences together, and can even predict how a sentence will end.

To put this advance into a real-life situation, the Inserm researchers incorporated this new brain into the iCub humanoid robot.

To test it, researchers asked the iCub robot to point to a guitar, then asked it to move a violin to the left. Before performing the task, the robot repeated the sentence and explained that it had fully understood what it had been asked to do.

The system can be used to improve understanding of how the brain processes language,” says CNRS Director of Research at Inserm Unit 846 Peter Ford Dominey.

“We know that when an unexpected word occurs in a sentence, the brain reacts in a particular way These reactions could hitherto be recorded by sensors placed on the scalp,” he says.

The model makes it possible to identify the source of these responses in the brain, and could contribute to possible linguistic malfunctions in Parkinson’s disease, he says.

But it also could enable robots to learn a language one day. “At present, engineers are simply unable to program all of the knowledge required in a robot,” says Dominey. “We now know that the way in which robots acquire their knowledge of the world could be partially achieved through a learning process – in the same way as children.”