Australian scientists have invented a new breed of robots called Lingodroids, programmed to make, use, and share language. The bots can coin words to describe places they have been, places they want to go, and plans for getting there. “When they need a new word, they invent one,” says Janet Wiles, a cognitive scientist at the University of Queensland who leads an interdisciplinary team on the project.
The rolling chatterboxes “see” using 360-degree cameras, laser range finders, and sonar. A microphone functions as their ears, and a speaker acts as a voice box, emitting the familiar beeps of a touch-tone phone. As for brains, Wiles outfitted each Lingodroid with an alphabet of beeps that correspond to letters. Then she programmed them to play a series of games in which they paired the letters into nonsensical combinations like “ja” or “ku” and joined those syllables to coin neologisms as needed. For example, in one game two robots roamed through a course and met in an unfamiliar part of it. The meeting triggered one robot to name the spot “jaya” and share the new word with its partner, who then added the word to its lexicon. In this way the robots slowly built a new language to describe their travels [pdf] and eventually even learned to communicate and understand directions.

Wiles notes that although the language may seem simple, for robots, grasping spatial information is incredibly complex. “We don’t realize how sophisticated our use of language to describe the world around us is,” she says. Ultimately, she hopes to teach her robots to chat up humans, paving the way for robotic caregivers, companions, and butlers.