Episode 622: Humans vs. Robots : Planet Money If you aren't already worried about being replaced by a robot, maybe you should be. Today on the show, three races pit humans against machines.
NPR logo

Episode 622: Humans vs. Robots

  • Download
  • <iframe src="https://www.npr.org/player/embed/405270046/405348208" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
Episode 622: Humans vs. Robots

Episode 622: Humans vs. Robots

  • Download
  • <iframe src="https://www.npr.org/player/embed/405270046/405348208" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

STEVE HENN, HOST:

Remember the folk tale of John Henry? He was a steel driving man a hundred years ago in the mountains of West Virginia, racing a machine to dig a tunnel through a mountain. John Henry won that race, but at the end, his heart gave out and he died. And that's where the legend of John Henry ends.

What the legend doesn't really dwell on, of course, is that in the end, the machines won. I mean, it wasn't just that he died; machines took over all those jobs building the railroads, just like they've taken over so many other jobs over the last 100 years.

STACEY VANEK SMITH, HOST:

In the '80s, there was probably a John Henry of auto manufacturing.

HENN: Beaten by one of those huge industrial robots.

VANEK SMITH: In the '90s, there was probably a John Henry of travel agents.

HENN: Taken down by Travelocity, wiped out by Expedia.

VANEK SMITH: And soon there will probably be a John Henry of taxi drivers racing against the autonomous car. And machines will probably win.

(SOUNDBITE OF SONG, "FORGIVEN NOT FORGOTTEN")

HENN: Hello, and welcome to PLANET MONEY. I'm Steve Henn.

VANEK SMITH: And I'm Stacey Vanek Smith. Today, so many of us are in a John Henry-style race against machines. So many of us are seeing part or all of our jobs taken over by robots - by software.

HENN: There are robot financial advisors, robot radiologists. And today on the show, we have three stories from the frontier, three races that pit humans against computers, including one where a computer has challenged a radio reporter to a dual.

VANEK SMITH: One of our own.

(SOUNDBITE OF SPONSORSHIP)

UNIDENTIFIED WOMAN #1: Support for PLANET MONEY and this message comes from Dropbox, maker of Dropbox for Business. The average person regularly collaborates on five or more different file types. Dropbox for Business lets you work with team members on any file, from Word docs to PDFs to Photoshop images and large videos. Share files of any size. Access them on phones and tablets, and gather feedback from your team. Dropbox for Business also features admin controls to keep collaboration secure. Learn more at dropbox.com/npr.

VANEK SMITH: For our first robot race, we wanted to start with something really simple, something anyone can do. Something we do all the time every day.

PIETER ABBEEL: Folding a towel, right. Just a towel.

HENN: Pieter Abbeel is a computer scientist at Berkeley. And seven years ago, he set out on this quest. He wanted to teach a computer - a robot, really - to fold laundry.

VANEK SMITH: That's genius.

HENN: (Laughter). Right? Who wouldn't want a robot like that? If you imagine like a washer and a dryer - that third box, a folder. I live for that box.

So I went up to Berkeley to visit Pieter Abbeel and his laundry folding robot. And I kind of expected a box. But when I got there, his robot wasn't a box at all. It was this little bot that looked kind of like a person. It had a head with two big eyes, a prominent chin. He reminded me of ET. And like ET, he could stretch and get taller. He starts at like 5'2", but he can get, I don't know, to be like 5'10". He has two arms with clamps on the end, perfect for picking up laundry. And Peter said they even gave him a name.

ABBEEL: The name is Brett, B-R-E-T-T. And it stands for the Berkeley Robot for the Elimination of Tedious Tasks.

HENN: To Pieter Abbeel, Brett is kind of perfect. He just wanted to teach him to do this simple task - to teach him to fold laundry. I mean, how hard could it be? Turns out, it's really hard because in every laundry basket, there is complete chaos. So Pieter started with the most basic thing - trying to teach Brett to fold a rectangular towel. They programmed him to find the corners of the towel and put the ends together. But Brett just kept poking at the pile of towels.

(Imitating robot voice) That's not a corner. That's not a corner.

To a computer, it's hard to see where the underwear stops and a towel begins. Every pile of laundry is different. And Peter showed me this picture of Brett just standing there looking confused.

So it looks like Brett has failed at finding the corner there.

ABBEEL: It's not figured it out yet at this point here. (Laughter).

HENN: He looks so sad. You know, he's like looking down, his chin is tucked.

ABBEEL: It looks like he's trying and then maybe he doesn't succeed and we're just wondering like, oh no, what can we do to help him out?

HENN: Initially, they hoped they could create a program that would help Brett figure this all out on his own. But it just wasn't working. So the team spent months staring at laundry baskets, holding towels up in the air, taking pictures of laundry. The solution was super complicated.

ABBEEL: Can you use multiple images to build a 3-D model of the current shape? 'Cause once you can do that, then you can analyze that 3-D shape, find where the corners are. And so that's what effectively we spent a lot of our time on is figuring out that process.

HENN: Once the 3-D computerized corner recognition system was up and running, Brett could fold a towel.

ABBEEL: So our robot initially would take about 20 minutes per towel.

HENN: What? Twenty minutes per towel?

ABBEEL: So yes, 20 minutes.

HENN: Yeah, I actually - before I came, I looked up a - I was looking for examples of things folding towels, and I found one of a toddler. And if you just give me a second.

(LAUGHTER)

HENN: OK, so she looks - how old do you think she is? Maybe 18 months old?

ABBEEL: I would say so, yeah. About 18 months old. She's...

(SOUNDBITE OF ARCHIVED RECORDING)

UNIDENTIFIED CHILD: (Crying).

UNIDENTIFIED WOMAN #2: You can do it. Again.

ABBEEL: Trying to learn to fold this towel.

(SOUNDBITE OF ARCHIVED RECORDING)

UNIDENTIFIED WOMAN #2: She's folding a towel.

HENN: Look, she grabbed the corner.

ABBEEL: She did find the corner pretty quickly there.

HENN: Oh, look at that. Look how fast that was.

(SOUNDBITE OF ARCHIVED RECORDING)

UNIDENTIFIED CHILD: I did it.

UNIDENTIFIED WOMAN #2: All right. Again.

HENN: That took seven seconds.

(LAUGHTER)

ABBEEL: Impressive.

HENN: And she's 18-months-old. How long did you work on teaching your robot?

ABBEEL: Two years.

(LAUGHTER)

ABBEEL: But two years of several students spending a lot of time on it.

HENN: Now, to be fair to Brett, after a couple more years of work, they had him folding a towel in just about a minute and a half. After a few more years, he was able to sort laundry. He could tell the difference between a towel and a pair of socks or a shirt. And he could fold them all. At his best, Brett was able to fold an entire small load of laundry in just about an hour.

ABBEEL: I think that's one of the most surprising things. Once you start working in robotics, you realize that the things that kids learn to do - maybe up to age ten - would happen - whatever we learn before age ten are actually the hardest things to get a robot to do. Whatever we learn later in life turns out easier to build artificial intelligence for. For example, playing a game of chess - something pretty much nobody would master before the age of ten, or definitely not be a world champion. But a computer is able to do it.

HENN: But you ask a computer to reach into a pile of Legos and stack blocks, and it's likely to get stumped. That's the thing about machines and computers, they thrive in situations where there are rules, where there's a clear structure. And as we're racing against machines in all of our lives, you should step back for a second and ask yourself, is my job structured? Are there rules that define what I do? Or is it more like a big, messy pile of laundry?

(SOUNDBITE OF SONG, "STARTED FROM THE BOTTOM")

DRAKE: (Singing) Started from the bottom now we're here. Starting from the bottom now the whole team here.

VANEK SMITH: OK, Steve. So so far in our raise - humans versus machines - robots zero, humans one.

HENN: Toddler - one.

VANEK SMITH: For our next race, robot versus therapist.

ELLIE: Hi, I'm Ellie. Thanks for coming in today.

VANEK SMITH: As robot overlords go, Ellie looks pretty nice actually. She kind of looks like a librarian inside of a videogame. She has brown hair pulled back in a loose bun. She's wearing a little beige cardigan. And she lives at USC's Institute for Creative Technologies. Skip Rizzo is one of her developers. And he got his start as a psychologist specializing in people with brain injuries. And rehab for these kinds of injuries can be really hard. It's really tedious. And some people really struggle with it. There was one guy, he was in his early 20s. He'd been in a car accident. And Skip said he was just getting really, really frustrated. It was hard to get him to do the rehab.

SKIP RIZZO: One day I see him sitting under a tree hunched over some. And I walk over and I go, Tim, what do you got? He goes, it's a new thing. It's a Game Boy. And he was playing Tetris. And he was essentially become a Tetris warlord. And he was glued to it. He was just enamored with it. And he was getting better at it. And that's when the light bulb went off.

VANEK SMITH: What Skip realized is that in some cases, it's easier for people to interact with a computer than it is for them to interact with other people. And Skip thought this might also be true for people suffering from emotional injuries, especially people who have trouble talking about their experiences. So Skip teamed up with the military to develop Ellie. And she is designed to work with people who are suffering from depression and posttraumatic stress disorder.

ELLIE: What's something you feel guilty about?

VANEK SMITH: OK, umm...

But there were some problems with Ellie. Getting a cartoon on a screen to have a conversation is really, really hard. Conversations are messy, a lot like a laundry basket. And at first, Ellie would interrupt people when they paused in the middle of an answer.

ELLIE: That's great.

VANEK SMITH: But then, they figured something out. Skip realized that Ellie didn't really need to understand what people said.

ELLIE: Tell me about the hardest decision you've ever had to make.

VANEK SMITH: (Laughter). OK, wow. You just like really like go right into it, huh? OK. The hardest decision I've ever had to make?

OK, you don't need to hear my answer. And actually my answer doesn't matter. It's the way I say it that matters. I'm in a booth and Ellie's up there on the screen and there's a microphone recording my responses, analyzing my tone and my pauses. And there are these big, expensive cameras that are tracking all of my expressions, how I move my eyebrows, where I look, my smiles.

RIZZO: Contrary to popular belief, depressed people smile as many times as non-depressed people, but it's almost like polite smiles rather than, you know, real, robust coming from your inner soul kind of a smile.

VANEK SMITH: So Ellie is collecting data on my smiles and all of my other expressions and micro expressions. And I actually saw a video of myself as Ellie sees me. And it was so crazy. She'd projected this weird blue grid all over my face that had outlines my mouth and my eyes and my eyebrows. And she was tracking everything I did. Ellie has figured out rules for emotions, or at least the physical expression of them.

ELLIE: Tell me about your relationship with your family.

VANEK SMITH: OK. I have a good relationship with them. I love them.

And this technique works. In a series of studies, Ellie was as good, or in some cases, even slightly better than psychologists at diagnosing PTSD and depression. And Skip says this makes some sense to him, because after all, humans can be fooled if someone looks really together or they're really charming or they spin a great tale. But Ellie isn't affected by any of that. And surprisingly, soldiers didn't mind talking about their experiences with a robot. In fact, they liked it.

JODY MITIC: Ellie - well, Ellie is a strange girl but I like her.

VANEK SMITH: This is Jody Mitic. He served with the Canadian forces in Afghanistan. He lost both of his feet to a bomb. And he had a session with Ellie.

MITIC: I'll be pretty frank. I walked into the room thinking I would be talking to a cartoon and I was going to have a bit of fun. And I went in the room, and Ellie started talking to me. Within no time, I think the question she asked me was is there anything you regret from your time in Afghanistan? And I broke into a story that I was driving in a convoy through Kabul and there was a woman in the middle of the road begging for money, and she had an infant in her arms that looked dead to me. And I've always regretted that I didn't stop the convoy and - just to check on her. I regret not taking some kind of action at that moment. And she got it out of me without even trying.

VANEK SMITH: Why do you think she did get that out of you?

MITIC: I don't know. I think it could be something about the fact that there's no judgment in her eyes. And there's no reaction no matter what you tell her.

VANEK SMITH: Our humanness can get in the way of doing this job. I mean, even a trained therapist can be shocked or judgmental. And even if they don't feel those things, the possibility that they might have that very human reaction, that has an effect.

Do you think that a computer has an edge over a person in this particular case?

MITIC: Yeah because, you know, a lot of therapists, you can see it in their eyes when you start talking about some of the grizzly details about some of stuff you might've seen or done. They are having a reaction, whereas Ellie seemed to just be listening.

VANEK SMITH: Well, there you have it. When it comes to diagnosing and working with people who have PTSD, it looks like robots might have an edge over people.

HENN: Yeah, I think we have to give that one to the robot.

VANEK SMITH: We're tied up.

HENN: Robots, one. Humans, one.

VANEK SMITH: We need a tiebreaker.

(SOUNDBITE OF SONG, "IF I COULD CHANGE YOUR MIND")

HAIM: (Singing) If I could change your mind, I would keep right on running.

HENN: As we were researching the story, we noticed something. There are robots that are beginning to do our jobs. I mean, not hosting podcasts really, but writing business stories - simple, formulaic stories that you hear all the time, like this one.

(SOUNDBITE OF ARCHIVED BROADCAST)

LAKSHMI SINGH, BYLINE: The Dow is up nearly half percent at 17,924. NASDAQ got more than half a percent at 49, 46 and the S&P 500 up slightly at 2,088. I'm Lakshmi Singh, NPR News.

HENN: That was probably written by a production assistant, or Lakshmi Singh just ad-libbed it looking at numbers in the studio. But it doesn't have to be written by a human. Stories like that are being written by robots all the time now. There's a company down in North Carolina called AI, Automated Insights. Its CEO is Rob Allen.

ROB ALLEN: So Automated Insights has a platform called Wordsmith that takes data and turns it into stories that sound as if a person wrote it.

VANEK SMITH: It's like a robot that writes news stories.

ALLEN: Something like that, yeah.

HENN: Wordsmith is getting a lot of work. It's writing sports stories for Yahoo! News. It's writing earnings reports for the AP. So we have our robot - Wordsmith. Now we need our human being, our John Henry. And we picked the fastest news reporter we know.

SCOTT HORSLEY, BYLINE: Scott Horsley, NPR News, the White House.

HENN: Scott Horsley. Inside NPR, this guy is a legend. I asked him how he got so fast, and he said he got his start at a commercial news station.

HORSLEY: I was kind of like this machine. I mean, I was just churning out copy hour by hour. And the expectation was you would write 10 readers an hour so...

VANEK SMITH: Whoa.

HORSLEY: ...The average would be, you know, less than five minutes.

VANEK SMITH: If anyone can beat the machine, Scott Horsley can.

HENN: Here are the rules of the race. We were going to pit Scott against the machine writing a story about an earnings report. Earnings reports come out every quarter. Every public company has to file one. So we picked Denny's, the restaurant chain. When Denny's filed its earnings, after the market closed at 4 o'clock, Scott would grab the report, the computer would get a copy, and they'd be off. Scott felt like he might have a little bit of an advantage because he was a Denny's regular. His waitress, Genevieve, even knew his name.

HORSLEY: When I used to go there, it got, you know, it got to the point where Genevieve would say, you just want the regular? And I could say, yeah.

VANEK SMITH: Oh.

HENN: Oh, that's good.

HORSLEY: It was good because I didn't have to say out loud moons over my hammy.

VANEK SMITH: (Laughter).

HENN: All right, it's after 4 o'clock. The market is closed. Let's do this thing. Man versus computer.

VANEK SMITH: OK.

HORSLEY: On my marks.

VANEK SMITH: OK. Get set. Go. Horsley is typing, right away starts typing so fast.

HORSLEY: Oh, Scott Horsley was a news spot writing man.

HENN: In North Carolina on the Wordsmith side, the people writing the computer were pretty relaxed.

So the race is officially on, but go ahead.

UNIDENTIFIED MAN: OK, exciting. I will tell my engineers that the race is officially on because they've gotten the earnings report. He's smiling now.

HENN: It sounded like a hubbub of activity behind you when you said that.

UNIDENTIFIED MAN: Yes. Although that is in this case probably people getting ready to play ping-pong or foosball or something like that.

HENN: These guys don't have to do anything. It all just happens. And before the race, they had bragged that Wordsmith was pretty much instantaneous. But now, they're telling me there may be a small delay. They have to wait for a database to update before the story will be ready. So Scott actually might have a chance.

VANEK SMITH: And Scott is writing like the wind.

HORSLEY: OK, the company is also forecasting better growth in the coming year, which I didn't copy the data over here.

VANEK SMITH: I think the computer just finished.

HORSLEY: Well, all right.

HENN: After a few more minutes of furious typing...

HORSLEY: All right, ding.

VANEK SMITH: All right, seven minutes and 30 seconds.

HENN: And I just have to say, I have never written a news spot that fast. Never in my life have I finished in seven minutes and 30 seconds. I mean, anyone in this business is just bowing down to the glory of Horsley.

VANEK SMITH: It's so true. Everybody, of course, except for one reporter, Wordsmith.

HORSLEY: And the computer took how long? Two minutes?

VANEK SMITH: Two minutes.

HORSLEY: (Laughter).

VANEK SMITH: But this isn't only about speed. It comes down to content now. We presented each of the spots to NPR's newscaster Lakshmi Singh, and she recorded them.

HENN: OK so this is going to be spot number one. It's the first time I've ever heard it. I don't know if it's a computer or human. Ask yourself if you can tell the difference.

(SOUNDBITE OF ARCHIVED RECORDING)

SINGH: Denny's Corporation on Monday reported first-quarter profit of 8.5 million dollars. The Spartanburg, S.C.-based company said it had profit of ten cents per share. The results beat Wall Street expectations. The average estimate of four analysts surveyed by Zacks Investment Research was for earnings of nine cents per share. The restaurant operator posted revenue of 120.2 million in the PRA also beating Street Forecast.

HENN: OK, I think that was a computer. It was pretty straight-ahead. It was just the facts. So I'm going to say that was our robot.

VANEK SMITH: You think that was the robot?

HENN: I do.

VANEK SMITH: OK. Well, we have another spot to play for you. This is spot number two. So got to make the call. Computer or human?

(SOUNDBITE OF ARCHIVED RECORDING)

SINGH: Denny's Corporation notched a grand slam of its own in the first quarter, earning a better-than-expected 10 cents a share as restaurant sales jumped by more than 7 percent. Operating revenues topped $120 million. Adjusted net income jumped 36 percent to 8.7 million. Denny's is one of the nation's largest full-service restaurant chains. The growth in sales suggests consumers are opening their pocketbooks for pancakes, eggs and hash browns.

HENN: OK, I'm thinking that's Scott. There was just more breakfast.

VANEK SMITH: (Laughter). You are absolutely right, Steve Henn. In fact, number two spot was our human, Scott Horsley. It looks like his time with Genevieve paid off. But you also have to keep in mind that the robot did beat Horsley time wise. So I'm going to let you call it, Steve. Who won this one, man or machine?

HENN: It's tough. Scott's piece was more fun to listen to, but, you know, the machine got the job done, and it got the job three times as fast.

VANEK SMITH: Yeah.

HENN: So as much as I hate to do this, Scott, I'm sorry. You lost.

VANEK SMITH: Well, Scott, of course, took it really well. And he did remind me that John Henry died at the end of his race.

HENN: It could have been worse.

VANEK SMITH: (Laughter). Exactly. He said he felt like he got off easy.

HORSLEY: You know that movie where the guy comes across the finish line like two days after everybody else has finished in the marathon? That's what this was like. But people still cheer for him because he's got so much heart. I'm going to be in the sympathy real - at the old reporters, and then there was Horsley. The race was over, most of the crowd had left when Horsley, limping and gasping for air, came into the stadium, crawling across the finish line. Automated Insights was home, feet up on the couch drinking whatever it is that computers drink after work.

VANEK SMITH: You see, and that's where the human touch comes in. Right there.

(LAUGHTER)

VANEK SMITH: Scott Horsley, thank you.

HORSLEY: It's my pleasure. And congratulations to our future computer overlord, Automated Insights. I'd like to be working for you one day, Mr. Computer.

HENN: So, Stacy, it looks like we better start learning to fold towels.

VANEK SMITH: It does.

(SOUNDBITE OF SONG, "THE LEGEND OF JOHN HENRY'S HAMMER")

JOHNNY CASH: John Henry hammered in the mountains. He'd give a grunt and give a groan with every swing.

HENN: If you think I got it wrong, if you think Scott Horsley kicked that computer's butt, you can email us at planetmoney@npr.org or send us a tweet on Twitter @planetmoney.

VANEK SMITH: #TeamHuman. We also have a couple people we'd like to thank. A special thanks to Jonathan Gratch and the whole team at USC who helped me talk to Ellie. That was great. Thanks you guys.

HENN: And thank you to our producer, Nadia Wilson, for putting this all together and putting up with us.

VANEK SMITH: And NPR recommends you listen to StoryCorps. The StoryCorps podcast collects unscripted stories about real life. Find the StoryCorps podcast now at npr.org/podcasts. I'm Stacy Vanek Smith.

HENN: I'm Steve Henn. Thanks for listening.

(SOUNDBITE OF SONG, "THE LEGEND OF JOHN HENRY'S HAMMER")

CASH: There was a big crowd of people at the mountain. John Henry said to the steam drill, how is you? Pardon me, Mr. Steam Drill, I suppose you didn't hear me. I said how are you? Well, can you turn the jack? Can you lay a track? Can you pick and shovel, too? Listen, this hammer swinger is talking to you.

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.