Skip to main content

Smarter Machines

A Google engineer plays ping-pong with a robot. The robot learns from its mistakes and fixes them. It learns quickly with artificial intelligence. MONICA RODMAN—THE WASHINGTON POST/GETTY IMAGES

Google is teaching robots to be more like people. Some can play ping-pong. They can fetch a bag of chips from the kitchen drawer. But here’s the amazing part: They learn how to do these things on their own.

At Google’s robotics labs, scientists are testing new AI technology. It allows robots to teach themselves how to do things.

Look at the robot playing ping-pong, above. Its camera takes in information about the game. The robot uses that information to learn how to hit the ball. It tries different moves. It gets rid of those that don’t work. It improves on those that do. And it becomes a better player.

A Google robot learns how to catch a ball. Like a person, the robot can get better at the task with practice.


Teaching robots to do things isn’t new. Robots build cars in factories. They move heavy boxes around in warehouses. But robots that can act on their own are new. Soon, one might be fixing you a snack at home. Who knows? You might even think of it as a member of the family.