How Can Robots Learn New Tasks? Practice, Practice, Practice : All Tech Considered Even the smartest robot does a miserable job picking up objects it hasn't been programmed to recognize. One way robots may get better at it is to learn by experience, a researcher says.
NPR logo

How Can Robots Learn New Tasks? Practice, Practice, Practice

  • Download
  • <iframe src="https://www.npr.org/player/embed/455507215/455657530" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
How Can Robots Learn New Tasks? Practice, Practice, Practice

How Can Robots Learn New Tasks? Practice, Practice, Practice

  • Download
  • <iframe src="https://www.npr.org/player/embed/455507215/455657530" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

KELLY MCEVERS, HOST:

Robots have been around for a while now. And you'd think by this point, they could do things that are useful to us normal people like put away the groceries or unload the dishwasher. But it turns out these tasks are actually hard for robots. Now a researcher at Brown University thinks she has a way to help robots get better at doing real-world things. As part of his series Joe's Big Idea, NPR's Joe Palca introduced us to a computer scientist who thinks robots will be more useful if they learn to do things for themselves.

JOE PALCA, BYLINE: I'd heard Brown had some hotshot computer robotics people, so I was expecting great things when I got there. In a lab in the Watson Center for Information Technology, computer scientist Stefanie Tellex introduced me to a popular industrial robot known as a Baxter. The Baxter is basically a boxy torso with bright-red arms and grippers where the hands would be. Tellex wants to show me what the Baxter can do.

STEFANIE TELLEX: So this a really, really cool project.

PALCA: She's strewn a bunch of different objects on the table were the Baxter is sitting. It's an eclectic collection - things like a 9-volt battery, a highlighter pen, a plastic kazoo. The Baxter's job is to pick an object up.

TELLEX: What it does, is, it picks it up, and it moves it up and down.

PALCA: So it can show it has a good grip on the object, and then it sets the object back down on the table. I watch the Baxter reach for the battery. Watching the Baxter do this is a little like watching paint dry. It seems to take forever for the robot to reach the battery, and when it does, it misses it with the gripper and just knocks it over. I watch for a while longer, and finally, I can't stop myself.

It looks pathetic.

(LAUGHTER)

PALCA: It's really, really pathetic that this is something that a sophisticated - which I presume this is...

TELLEX: It is, yes.

PALCA: ...Is spending any time doing.

TELLEX: A lot of the stuff that is really, really, really hard for robot to do is almost effortless for a person to do. But why is this a problem at all? Everybody can do that. My 2-year-old can do that.

PALCA: But Tellex says look at it from the robot's point of view. This isn't a factory where the robot has been programmed to do a very specific task. This Baxter doesn't know anything about batteries or kazoos. All it has is the information from its cameras. But that information is just a bunch of numbers.

TELLEX: It's basically this matrix, this big matrix of numbers, and it somehow has to look at that matrix of numbers and then run a program to figure out where the object is and what it is and where it should put its gripper in order to not be so pathetic and actually pick it up effectively.

PALCA: She thinks the way robots will get faster and smoother at picking up unfamiliar objects is to give them programs that let them learn from experience just like a child would. After robots pick up a battery or a kazoo a couple dozen times, it'll begin to recognize them reliably. But learning takes time, so Tellex has got her Baxter working round-the-clock picking up objects and putting them down, picking up objects and putting them down, picking them up and putting them down. And she has an idea for speeding up the learning curve.

TELLEX: So everybody either has a Baxter or has a friend with a Baxter in robotics research right now.

PALCA: Tellex says a lot of the time, those robots aren't being used.

TELLEX: At night, for example, you know, students go home and go to sleep, and they robot's just sitting there. And what it could be doing instead is collecting this type of data.

PALCA: Tellex is hoping to recruit some of these robots to do the same tasks as her Baxter to speed up the learning process, a kind of robotic version of many hands make light work sharing what they've learned. And who knows? Maybe someday there will be a Baxter to put away the groceries. Joe Palca, NPR News.

Copyright © 2015 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.