Scientists have been trying to figure out how kids pick up language so fast for decades, and a new study out of the Okinawa Institute of Science and Technology (OIST) might have cracked part of the puzzle: curiosity.
Researchers built a virtual robot with a brain-inspired neural network and set it loose in a simulated 3D world full of shapes, colors, and simple commands like “push left magenta dumbbell.”
Some robots were rewarded only for completing tasks correctly. Others got an extra reward for curiosity, essentially getting a little internal high whenever they encountered something that challenged their existing understanding of the world.
How does curiosity help a robot learn language?
The curious robots didn’t just edge out their indifferent counterparts; they blew past them. According to the study, published in Science Advances, curious robots reached a genuine understanding of language in about half the time.
Study author Theodore Tinker compared it to trying white chocolate for the first time even though you already love dark chocolate. You take the risk anyway, and you walk away knowing more about chocolate in general.
Things got even more interesting halfway through training. The curious robots started knocking things over and experimenting with actions nobody asked for, basically playing. Nobody programmed that behavior in. It just showed up on its own.
Do robots really make the same mistakes as toddlers?
The robots also mimicked a well-known quirk in how children learn language. Kids often get certain verb forms right at first, then start applying grammar rules too broadly and make mistakes on verbs they’d previously used correctly, before eventually sorting out the exceptions and correcting themselves. The robots followed the same U-shaped dip in performance.
It’s also a nice contrast to how today’s chatbots learn. Large language models like ChatGPT train on massive datasets and spit out the statistically likely next word. This robot’s brain works more like ours, prioritizing accuracy while trying to keep its beliefs intact, only updating them when something surprises it enough to be worth the trouble.
None of this means robots understand language the way we do. But it does suggest that curiosity paired with a wide variety of experiences might be a big part of how toddlers crack the language code with so little to go on.