Cornell team creates robot that understands casual commands, makes dessert (video)

Cornell researchers have built a robot that can understand natural language commands.

The Robot Learning Lab released two videos showing a robot translating casual commands to make Top Ramen and an ice cream sundae. The first video reveals how the robot translates language into logical connectives, and the second shows an ice-cream-serving version that’s sure to the make the powerful fro-yo union go cray cray.

The researchers have begun to crowdsource the intelligence of their algorithm with a website that asks users to make their own commands for the robot.

“The idea of using crowdsourcing — where volunteers can generate such step sequences by playing a video game — is very creative,” says Stanford artificial intelligence expert and chief scientist at Baidu, Andrew Ng.

But Ng says we won’t be seeing this kind of robot at Froyo stores any time soon — or in homes.

“We’re many years away from having robots that can do general household tasks. (More than 5 years.) For example, almost every home’s kitchen is different, and it is difficult to program robots to be robust enough to work in any kitchen. But understanding natural language is an importance piece of the puzzle.”

You can play the simulator here.