Demo: Teachable Robots
Like any proud parent, Michigan State University computer scientist Juyang Weng has a lot to say about what sets his little ones apart from their peers. Traditional robots, he explains, must be specially programmed for new tasks. And you just can’t teach them much. Sure, they can acquire data-but only within narrowly defined parameters set ahead of time by their programmers. “But human learning is not like that,” Weng says. “Human learning is real-time, online, on the fly.” And that kind of learning, Weng says, is essential if you want a machine to be able to cope with the unexpected-unpredictable terrain, new people or objects, noisy settings-which will surely confront robotic household assistants and military machines alike.
In 1994, Weng and his team set out to build a robot with a capacity for learning like that of a human baby. They came up with a black, moon-faced machine named SAIL, short for Self-Organizing Autonomous Incremental Learner, endowed with what Weng calls a “developmental program”-a program that imparts attributes such as curiosity. Then SAIL was “born.” “Birth’ means that the robot starts to interact with the real world, just like a baby interacts with his doctors, his father, his mother,” Weng explains. “These interactions make the robot gain a sense of the outside world.” Through such exploration, SAIL has learned tasks like navigation, identifying and sorting objects, even some speech. And he now has a younger-though physically more sophisticated-sibling, Dav. Weng introduced his robotic family to Technology Review senior editor Rebecca Zacks.
Keep Reading
Most Popular
Large language models can do jaw-dropping things. But nobody knows exactly why.
And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.
How scientists traced a mysterious covid case back to six toilets
When wastewater surveillance turns into a hunt for a single infected individual, the ethics get tricky.
The problem with plug-in hybrids? Their drivers.
Plug-in hybrids are often sold as a transition to EVs, but new data from Europe shows we’re still underestimating the emissions they produce.
Google DeepMind’s new generative model makes Super Mario–like games from scratch
Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.
Stay connected
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.