Complex systems scientist Janet Wiles works with robots that create their own language — without her interaction.
To do this work, iRats, the linguistically-gifted robots that Wiles studies, are let into a controlled landscape and given the chance to explore. In 20 minutes, she says, the iRats “link[s] the places it’s been into paths and then paths into a map.”
Their personal maps, drawn from the iRats’ own experience of the environment, allow the robots to create language and communicate. When robots meet in a mapped environment, they can communicate based on their experience of the environment. “When they meet each other, they stop and have a conversation called, ‘Where are we?’” Wiles says. “They create a word [to identify that place] … and then they go off exploring again.”
The robots repeat this process again and again, creating place names for points within an environment that allows them to further their language, creating terms for distances between two named points, even terms for imagined points they haven’t experienced yet.
And as their experience of space grows, so does their knowledge of time. “We have them notice the sunlight,” Wiles says, and as robots track brightness, they create language for time determinant on light. They create words for morning, afternoon, night and so on, and then learn how to coordinate meetings at specific places and time.
“Why are they exploring space and time?” Wiles asks. “It’s because [this is] the foundation of cognition.”
For more, watch Wiles’s whole talk below:
Insight from the TEDx office — why we like this talk:
The speaker is a researcher doing novel work with robotics and linguistics at a research university. She delivers a talk on a new technology — bio-inspired robots with linguistic capabilities — that is clear, interesting and supported by concrete examples. She supports her talk with clear images and video and breaks down a complex subject into an understandable narrative.