When It’s Not Haunting Your Nightmares, This Baby Robot Is Teaching Scientists About How Real Babies Learn

By
Photo: WIREs Cognitive Science

More than half a century ago, the computer scientist Alan Turing argued that babies were a lot like untrained robots: “Instead of trying to produce a programme to simulate the adult mind, why not rather try to produce one which simulates the child’s?,” he wrote in a 1950 paper in the journal Mind. “If this were then subjected to an appropriate course of education one would obtain the adult brain. Presumably the child brain is something like a notebook as one buys it from the stationer’s. Rather little mechanism, and lots of blank sheets.”

Several decades later, some robotics researchers have started to follow his lead. In a recent themed issue of the journal WIREs Cognitive Science focused on human development, Pierre-Yves Oudeyer, the research director at the French Institute for Research in Computer Science, made the case for baby robots as a means of helping scientists understand the real thing. “Researchers can actually build baby robots with mechanisms that model aspects of the infant brain and body, and then alter these models systematically,” he wrote, and “compare the behavior we observe with the mechanisms inside.”

In his paper, titled “What Do We Learn About Development from Baby Robots?”, Oudeyer highlighted a few ways that could happen. For example, the robot in the image above, called iCub, was built to model the cognitive capabilities of a toddler; other humanlike robots have been used to examine how the brain and body work together in the process of learning to walk.

And in a study that Oudeyer nicknamed “The Playground Experiment,” he and his colleagues built robots that taught themselves through trial and error to interact with several different objects (you can watch a video of the experiment here — the robots look like dogs, not babies, but the principle is the same). “Here a robot learns by conducting experiments,” he wrote in WIREs:

It tries actions, observes effects, and detects regularities between these actions and their effects. This allows it to make predictions. The way the robot chooses actions is like a little scientist: it chooses experiments that can improve its own predictions, which can provide new information, which makes learning progress, while continuously allocating some proportion of time to exploring other activities in a search for new possibilities.

Baby robots: not so great as a deterrent against teen pregnancy, fairly useful in the study of how we grow from infants to adults. They may look creepy, but at least they’ve found their niche.

This Baby Robot Is Teaching Scientists How Real Babies Learn