Season 1 • Episode 14
By the year 2000, the internet was already becoming a cesspool. The bad guys used software bots to sign up for millions of fake email accounts—for sending out spam.
PhD student Luis Von Ahn stopped them. He invented the CAPTCHA, that website login test where you have to decipher the distorted image of a word. Or you have to find the traffic lights or fire hydrants in a grid of nine blurry photos.
Those tests help to keep down the volume of spam, spyware, and misinformation; they advance the clarity of digitized books and the intelligence of self-driving cars; and, by the way, they made a handsome profit.
The only problem: We HATE those tests!
Guest: Luis Von Ahn, co-inventor of CAPTCHA, co-inventor and CEO of Duolingo.
Episode transcript
Captcha Script
Description
By the year 2000, the internet was already becoming a cesspool. The bad guys used software bots to sign up for millions of fake email accounts—for sending out spam.
PhD student Luis Von Ahn stopped them. He invented the CAPTCHA, that website login test where you have to decipher the distorted image of a word. Or you have to find the traffic lights or fire hydrants in a grid of nine blurry photos.
Those tests help to keep down the volume of spam, spyware, and misinformation; they advance the clarity of digitized books and the intelligence of self-driving cars; and, by the way, they made a handsome profit.
The only problem: We HATE those tests!
Guest: Luis Von Ahn, co-inventor of CAPTCHA, co-inventor and CEO of Duolingo.
Intro
By the year 2000, the internet was already becoming a cesspool. Software bots were signing up for millions of fake email accounts—for sending out spam.
Luis Von Ahn stopped them. He invented the CAPTCHA, that website login test where you have to decipher the distorted image of a word. Or you have to find the traffic lights in a grid of nine blurry photos.
The only problem: We hate that test.
LUIS: I would be at a party, and people would ask me what I did, And I would tell them that I helped invent that thing, and people would tell me, “Oh, I hate you.”
I’m David Pogue, and this is “Unsung Science.”
First Ad
Season 1, Episode 14: The Man Who Stopped the Spammers.
[music]
In his 43 years on this earth so far, Luis Von Ahn has had three ingenious, innovative, world-changing ideas. I guarantee that you’ve encountered his second one. Probably hundreds of times, actually.
Most of us have zero world-changing ideas. Occasionally, somebody has one. But three times?
His first idea came to him in Guatemala, where he grew up.
LUIS: I wanted to start a gym, where instead of charging people to—to show up, let people just show up for free. we were going to connect all the machines up to kind of the power grid, and we were going to use what—you know, the kinetic energy that people did or had whenever they were, you know, exercising, to generate power. And I thought we could make a lot of money from that.
Now, you will note that I did not say that all three of his world-changing ideas actually succeeded in changing the world.
Luis: I thought I was the first person to have this idea. Turns out, this is a very old idea. It also turns out it doesn’t work.
That’s right: The pedal-power gym idea flopped.
LUIS: (cont’d) It turns out, this is, this is not a good idea for many reasons. The biggest one of which is that humans are just not very good at creating energy.
DAVID: Oh!
LUIS: You just, just don’t make a lot of money from this. There’s another reason why that doesn’t work a lot. Turns out gyms make most of their money from people who don’t show up.
David: Of course.
Luis: Here, you kind of need people to show up.
To be fair, he was pretty new at the game when he had this first idea.
David: And how old were you at this point?
Luis: Twelve years old, eleven years old.
Things started going better six years later, when he came to the United States to attend Duke University. As the year 2000 dawned, Luis was at Carnegie Mellon, in his first year of working toward a PhD in computer science. And one fateful day, he went to a talk by an Israeli computer scientist named Udi Manber, who at this point was the chief scientist at Yahoo.
Luis: By the way, in the year 2000, Yahoo was the biggest name, you know, biggest tech company. In other words, like the Google of today. And you know, and he was giving a talk about ten problems that they didn’t know how to solve inside, inside the company.
And one of those ten problems that the greatest minds at Yahoo could not solve—was automated software spambots signing up for free Yahoo Mail accounts by the millions.
LUIS: (cont’d) Yahoo gave out free e-mail accounts. And there were people who wanted to send spam from Yahoo accounts, but each Yahoo account only allowed—only allowed you to send like 500 messages a day. If you wanted to send millions of emails, spam emails per day, then what these people did is they wrote programs to obtain millions of Yahoo accounts every day. And they didn’t know how to solve that problem, how to stop that.
So I started talking about it with a person who had just become my Ph.D. adviser. His name was Manuel Blum, or is Manuel Blum. He is still, most definitely still alive.
And, you know, we started thinking—and this is where this idea of a CAPTCHA came up.
The idea was this: Any time you tried to sign up for a Yahoo Mail account, you’d encounter a little puzzle. Something easy for a person to solve, but hard for a spambot.
LUIS: (con’td) the way to stop these spammers was to have a test that can distinguish between whether you’re a human or a computer. If you are human, then presumably you can’t get millions of email accounts because you get bored. Whereas if you’re a computer, you can get a million.
And so if the only entities that we we’re giving email accounts to were humans, then that would stop the spam.
CAPTCHA, the name he gave his online mini puzzle, stands for ‘Completely Automated Public Turing tests to tell Computers and Humans Apart.’ More or less.
Not sure if you’ve heard of the Turing test, but it’s incredibly famous among computer scientists. It’s this experiment proposed by British mathematician and computer scientist Alan Turing, who’s known as the father of artificial intelligence. There was actually a movie about him called “The Imitation Game,” where Benedict Cumberbatch played Alan Turing.
Clip from the movie: ”Do you want to play?…”
Anyway, the Turing test is intended to set a standard for determining if a computer has achieved true artificial intelligence.
Luis: When can we tell that a computer is actually intelligent? This is kind of like a philosophical test that said, like, “look, we’re going to have a human judge ask questions to two entities. One is the computer, one is the human.”
The computer and the human are hidden behind two curtains; the judge can’t see them. The judge types in questions, and looks at the text of the responses. If it’s impossible to tell which answer came from a person and which from the computer, the computer has passed the Turing test.
Luis: The judge can just ask whatever questions they want. And if we really can’t distinguish, then we’ll say that computer is really intelligent.
To this day, we have not made a computer that can actually pass the Turing test successfully. It’s just—it’s just too hard.
The funny thing is, if you really think about it, the Captcha problem is the opposite of the Turing test. The Turing test is successful if the judge can’t tell the difference between a person and a machine. The whole point of Luis Von Ahn’s project was to create a test that can tell the difference.
There’s another difference between the two tests, too.
Luis: (con’t) here’s the key. In this case, the judge was a human. In our case for the CAPTCHA, what we needed to do is we needed the judge to be a computer, because we need—we need the computer to determine whether it’s talking to a human or computer, which is, which is much harder, in some sense, at least for to grade it.
So, so I think the hardest thing was just coming up with this general idea that, like, “OK, what we need is a test that can distinguish humans from computers, but that computers need to be able to grade.”
Then after that, we started coming up with, like, “OK, what are things the computers are not very good at?”
In the year 2000, the answer was obvious: Computers are not very good at identifying what’s in pictures.
Luis: We quickly honed in on images, and just doing images of text, images of flowers, images of stuff. And then, and then after a while, the images of text were the ones that seemed like the best idea.
And then I just went and developed a program that, that distorted random text. And, and that was the first version of a CAPTCHA.
That’s right: The test they came up with presents you with the image of a typed word—but the letters are all, like, twisted, bent, and distorted, as though the typist were severely drunk and typing on Saran wrap. You are supposed to interpret what that word is, and type it into a box on the website.
SFX: computer keyboard typing
Actually, computers in the early 2000s were pretty good at OCR—that’s optical character recognition, meaning looking at a picture of text and figuring out what the letters are. But the added challenge of the twisty distortion really threw those OCR programs off the track.
David And behind the scenes, I mean, what is it? I mean, there’s got to be some, I don’t know, SQL database or massive bank of little images. I mean—
Luis Actually, there was no database at first. we would just write a program that, what it would do is it would place—it would pick some, first, random characters. We’d put them on an image, then it would distort them, and then we would save that image. And then we just had a couple million of those saved, not even in a SQL database, just they were there. So saved as files.
It worked brilliantly. The spambots didn’t have a chance.
At the time, Von Ahn had no idea if his invention would be of any commercial use. But one guy, he knew would be interested: Udi Manber, that Yahoo chief scientist, who’d given the talk that started this whole affair.
Luis (cont’d) We sent them an email saying, “Hey, we think we can solve your problem.” And he said, “Oh, that seems like it solves the problem.” And then, in fact, pretty soon after that, it was being used by Yahoo, and then basically, every website started using it, and there was just, you know, millions of websites out there were using it.
[music]
Well, how wonderful! Luis Von Ahn’s ingenuity 1, spammers zero. Internet saved!
Luis: [00:06:03] And, you know, at first I was very proud of myself, because, OK, look at the impact that my, my work has had. You know, basically, we stopped spam, it’s being used by a lot of people.
There was only one problem:
Music abruptly stops
People hated his invention.
LUIS: 0:00 How many of you had to fill out a web form where you’ve been asked to read a distorted sequence of characters like this? How many of you found it really really annoying?
OK, outstanding. So I invented that. (laughter)
That’s how he introduces himself in a 2011 TEDx talk at Carnegie Mellon.
Luis: I would be at a party, and people would ask me what I did, And I would tell them that I helped invent that thing, and people would tell me, “Oh, I hate you.”
That’s right: The inventor of Captcha is fully aware that people hate the thing.
LUIS: I say either— “well, I’m sorry.” Or,, “I find it annoying too.”
You heard it right here, folks: Even he finds them annoying!
In fact, Luis can tell you exactly how much of your time they waste.
Luis: I did a little back of the envelope calculation. At the time, about 200 million times a day, somebody typed one of these CAPTCHAs. 200 million times, times ten seconds, which is how long it takes to type one of these—you, humanity as a whole was wasting about 500,000 hours every day typing these annoying CAPTCHAs.
David Great.
Luis So I started feeling bad about that. And that’s when I started thinking, “OK, can we do something good with that time?”
See, the thing is kind of similar to the, the gym idea. Can we get millions of people to do something during that time that is actually valuable?
I’ll give you a hint: We’re only at the halfway point of this story. After the break, we’ll tell you what he came up with to make those half a million hours every day useful to humanity.
Break
Book Ad
Welcome back!
By 2005, Luis Von Ahn’s invention, the Captcha test, was a huge hit. It reduced the world’s scumbag spammers to blubbering losers. No longer could they bombard websites with phony signups, for the purpose of pursuing their pathetic spammy schemes.
Unfortunately, he had achieved this success by transferring the burden onto us. Treating us as though we were guilty until proven innocent. Now we were the ones being challenged. We were losing ten seconds per website, typing in those stupid distorted letters.
Now, to be fair, history is full of examples like that, where the actions of a few selfish, greedy idiots wind up inconveniencing billions of innocent people for the rest of our lives. Some dirtbag tries to put poison into drugstore Tylenol bottles, and now the rest of us are stuck with frustrating, plastic-wasteful bottle lids. Some delinquent tries to blow up a plane with a shoe bomb, and now we all have to walk through the TSA scanners in our socks.
Luis felt bad that his hacker blockade wasted everyone’s time, but at least he could do something about it.
Luis: So that’s very valuable time. So can we use it for something? And then we, we you know, I ended up coming up with this idea, that while you were typing a CAPTCHA, you could be helping digitize books. And here’s kind of how that works.
So at the time, the year, maybe 2005, 2006, there were a lot of projects trying to digitize all of the world’s books, where, where, you know, the way that worked is you start with a physical book, and you want to put it on the Internet. And the way you do that is you basically take a digital photograph of every page of the book.
Now, these are pictures of text. The next step in the process is that the computer needs to decipher what’s the text in there.
In other words, computers had to perform—c’mon, you know this term!—OCR. Optical character recognition.
Luis: (cont’d) And unfortunately for books that are older, where maybe the ink has faded, computers could not recognize many of the words.
At the time, computers couldn’t recognize about 30 percent of the words. So the thought, the idea was, let’s take all those words that the computers could not recognize while —while books are being digitized, and let’s get people to read them for us while they’re typing a CAPTCHA. what we started giving people were these words that the computer was not able to digitize, and or, to recognize.
So, yeah—all this time, you thought you were typing random words. In fact, you were helping companies digitize old books and articles. And, by the way, helping Luis’s little company make money.
Luis The idea is we made a CAPTCHA, a system, a whole system that would help your website be protected against spam. So, and we gave that away for free, on, for example, Facebook use our CAPTCHA, and we give it away for free. But always with the caveat—that if they are going to do that, then we can see the answers that users are typing so that we help to digitize something.
And the way we made money is by charging people who needed digitization stuff.
For example, The New York Times was our client. The New York Times had this old archive of old editions of The New York Times from 130 years of New York Times or something like that, from the, from the 1800s. And they, they needed this to help digitize their whole archive.
they were sending us all the scans they had scanned already, and we were sending them—we were taking all the words that computer could not recognize, and we were getting, you know, through the CAPTCHAs, people who are, for example, signing up for Facebook or Twitter or a lot of websites where you use CAPTCHA, they were helping us to digitize The New York Times. And we would make money from The New York Times.
(moved from earlier) It became very successful. And then Google bought it to help their Book Digitization whole project.
The new system, called reCAPTCHA, became an even bigger hit. Here’s how he described the aftermath in his TEDx talk:
LUIS: So every time you buy tickets on Ticketmaster, you help to digitize a book. Facebook: Every time you add a friend , you help to digitize a book. Twitter and about 350,000 other sites are all using reCAPTCHA. And the number of sites that are using reCAPTCHA is so high that the number of words we’re digitizing per day is really really large. It’s about 100 million a day, which is the equivalent of about two and a half million books a year. And this is all being done one word at a time by just people typing CAPTCHAs on the Internet.
(Applause)
David: There are some people who are a little nervous about Google being the owner of one of the most widely used CAPTCHA systems. I’m sure you’ve been asked about that.
Luis: Yeah, there are people who are nervous about that. I mean, I understand. I think, you know, this is these are very, very tricky questions. I mean, personally, I, I think the privacy fight is over. I mean, I, I I’ve given up on my privacy against large companies a while ago.
David: Wow.
Luis: And not only that, I also think— after having been inside Google, I saw with how much respect they treat user data, because they know that they are—you know, they know that, that they are a few scandals away from being in deep trouble. So they take it with— with a lot of care, I think.
David: And we should point out that Google has said, “we do not use data collected for advertising purposes.”
Luis: Yeah, that’s the case. And so I, and I actually believe them.
Now, remember: Luis said that the hard part was finding a test that was too hard for a computer to pass, but easy enough for a computer to judge whether the test had been passed. That’s been bugging me.
David: If the computer chooses a word that’s so distorted that it itself cannot do the OCR, then how does it know if we’re right?
Luis Yeah, that’s a great question. when we try to digitize books, this is what we do. We, we take a word that the computer does not know— We actually pair it with another word for which the computer does know the answer. And we actually give people both words, and we say, “please type both.” And we don’t tell them which one’s which, we just say, “hey, please type both.”
If they typed the word for which we know the answer, they type that one correctly, we assume that there are human, and we also get some confidence that they type the other word correctly. And then, and then what we do is, okay, so now we have a guess for what that other word is. We give it to like 10 other different people and we see if they type the same thing, and if they all type the same thing, we get with very high accuracy what that word really is. And that works.
One hallmark of the reCAPTCHA system, in other words, is that you have to type in two words.
Luis There’s sometimes also funny words that, um, funny combinations that happen, especially because we are showing two words at a time.
DAVID: Oh boy.
LUIS: I mean, there’s been all kinds of really funny examples, where it’s just like a website of a church, that’s says like, ‘Bad Christians.’ And it’s just—but these are just two randomly chosen words.
David : So we shouldn’t infer any evil on your part?
Luis: No, they’re random.
Now, a lot has happened since 2000, when Captcha came along, and since 2006, when you started unsuspectingly helping Google and the New York Times digitize their old pages.
Luis: You know, early on in the first version of the CAPTCHA, computers were pretty bad at recognizing distorted text, so they didn’t have to be that distorted. But over time, computers got better and better. And in fact, by now, computers are, in many cases, about as good as humans.
Because of that, we have to make them harder and harder.
David: A lot of times, the puzzles are so hard that even the human can’t pass the challenge. I’m sure you’ve been sent screenshots of words that are so munged no one can tell what it is.
Luis: Yes. That happens. I mean, it’s rare that that happens.
And that’s why the CAPTCHA itself, in true arms-race fashion, has evolved.
Luis: (cont’d) So what has happened is that for the more secure things, the CAPTCHAs have moved away from these distorted characters. And what is being used now are these—the puzzles are now things like, you see a bunch of pictures, and you have to click the ones that contain a stop sign—
David: Right, the traffic lights, the fire hydrants.
Luis: Yeah. It’s exactly the same idea as reCAPTCHA, except we’re not this time, we’re not trying to digitize books. This, a lot of times, comes from things like all the—all the mapping cars, or the self-driving cars. Basically, these are cars that are driving around that are capturing images of the whole world that are trying to figure out what’s around them. sometimes they cannot recognize what’s in an image. So it’s a similar case. It takes things—like, “is this is stop sign? I’m not sure. Okay, send it to human.” And then when you, when you get it, and you click on the stop sign, you’re actually helping either the self-driving car or the mapping software or whatever know that there is actually a stop sign, stop sign right here.
David: Oh. So we’re still doing good for the world as we do this.
Luis: Still doing good for the world.
David: Or for a company.
Luis: Or for a company. Just maybe not digitizing—or for a company, but maybe not digitizing books. but it’s a similar idea, it’s a thing that a computer cannot do.
David: You just solved a mystery for hundreds of millions of people—why it’s always traffic lights and fire hydrants we’re supposed to choose, and not bananas and puppies or….
Luis: It has to do with both self-driving cars and also mapping software.
OK. So now we kind of get why we have to put up with these challenges. Or we did 20 years ago! But really? Nothing better has come along since?
David: Are we sure that there’s nothing less annoying that we could do to thwart these spammers?
Luis: Yes, there is. by now, it did become a lot less annoying. I don’t know if you’ve seen that of late, where, you know, there’s a thing that says reCAPTCHA. We’re just trying to figure out whether you are human, and they just ask you to click somewhere. just, ‘click on this box’—that, that is much less annoying.
David: Sometimes, you don’t see anything except, ‘I’m not a robot.’
Luis: Yeah, yeah, yeah. I’m not a robot.
This is something that is, that that is done by Google. This actually comes from the original team. This is the company that they bought from me.
When you get that one, that means that in this particular case, probably means Google, has figured out that—yeah, you know what? We know you, because you’ve been around since 2016 in this computer, and you have a lot of Gmail emails, and you’ve done a lot of Google search queries. You’re a normal person, you’re not a spammer.
So they just do a little thing that just tries to double, double check that, that I can move the mouse or whatever. So, so one thing that has changed from the year 2005 to now is that there are companies like Google or like Facebook that, for the majority of people on the Internet, they kind of know who you are.
If you have a fresh computer that you’ve never used before, then you would have to do that annoying CAPTCHA. But for most of us, you are unlikely to have to type this as much as you were back in, say, the year 2005. It has become a lot better, you know, probably a little bit at the cost of your privacy.
OK, but wait a minute. We now know that computers eventually got too smart for the distorted-text-reading Turing test. Won’t they eventually get good enough to identify a few stupid stop signs in a photo grid?
Luis: It is, it’s a cat and mouse game. Now probably there’s a bunch of people working on making better recognition of stop signs or something like that. eventually— eventually computers are going to be able to do everything humans can. And so at some point, there won’t be a test that can distinguish humans from computers.
David: Well, wait a minute. Does that mean the end of the Internet? I mean, what happens is that—if there’s no sort of Turing test that works anymore?
Luis : I don’t think it’s the end of the internet, particularly because, like I said, more and more, these companies are going to know more and more about you. And I just don’t think there will be a humongous problem.
David: Okay, well, whatever the endgame is, why can’t we do that today? since we know it’s an arms race, since we know that eventually we’ll lose it to A.I. and computers, why can’t we jump to whatever will follow it now?
Luis: I’ll tell you why. this, by the way, is like 95 percent of the way there. I mean, really, for most of us, Facebook knows who we are. And Google knows who we are, so it’s 95 percent of the way there. The reason it’s not a hundred percent of the way there is because there are some people who really care about privacy. And, you know, there’s, there’s always going to be a kind of a way to browse privately. So, for example, there’s Chrome has private browsing. There’s all this stuff when people care about privacy.
I mean, there’s, there’s a tradeoff here, right?
David: Well, the irony is that it seems like most of the websites that present me with a CAPTCHA, I’m trying to get to in order to supply my name and address, like signing up for something.
Luis: Yes, it’s funny.
David: Why do I need privacy when the whole purpose is to supply my information?
Luis: Yeah, that’s funny.
Now, I mentioned at the beginning that Luis has had three world-changing ideas. You’ve heard about the gym membership that powers the grid, and you now know about Captcha.
But what about his third creation? It’s Duolingo, the language-training app. At this moment, it has half a billion registered users, learning 40 different languages, all for free.
Audio clip of DuoLingo in action.
And from the beginning, you could see the fingerprints of Luis Von Ahn, master of crowdsourcing, all over it.
Luis: In early Duolingo, as you were learning a language on Duolingo, you’re actually helping us to translate stuff that computers could not translate.
In fact, CNN was a client, so CNN would send us their news in English. We would then give it to people who were Spanish speakers who are learning English. And we would say, “Hey, you want to practice your English? Help us translate this CNN article into your native language of Spanish.” And so they would do it, and they would be learning English, and then we would get that translation, and then we would send it back to CNN and they would pay us for the translation. That was the very first version of Duolingo.
It turned out that, Just like the gym, it ends up being that you just —just can’t make much money from this.
And so we decided, “OK, just go, go to a business model where we actually give you ads.” And the way we make money is by showing you ads.
The dude just keeps doing that! He keeps coming up with ideas that make the world a better place, thwart the bad guys, and make a lot of money! It’s really a shame he gave up that electrical-grid gym thing.
David: Are there ever things that come to you in the shower that might be your —your big third act? I mean, honestly, to, to have the impact you’ve had, twice, is astonishing, but it makes me think there’s something in you that just has great ideas that can go really wide.
Luis: You know, as time passes, I am a lot more interested in literacy and teaching people how to read. I think with a computer, we should be able to teach the whole world how to read, significantly better than humans can teach you how to read.
You know, the U.S. is, the U.S. is fine. Most, most adults in the U.S. know how to read. But many countries in the world, there’s a significant fraction of people who don’t know how to read. In fact, there’s about a billion adults in the world that are illiterate. And I think we can—I think we can make a big dent with, you know, with a system to teach people how to read. So we’re working on that.
In the meantime, now you know why you have to encounter those infernal website challenges, you know how they came about, and you now consider them a necessary evil. Well, maybe you do.
David: Just for people who are like, “I don’t know what it is, I just don’t like doing it, I can’t even tell what’s a freaking traffic light.” Let’s just lay out what would happen if all of these challenges went away tomorrow. What would happen to the Internet?
Luis: Most likely, you would get a lot more spam in either your email spam, or you’d get a more kind of random Facebook followers that are not real people. these fake accounts can start boosting up bad political messages. There would be probably more fake news. They would probably be, you know, more spam.
David: Right. And from spam, phishing and spyware and….
Luis: Yeah, more spyware—the web would be a less safe place.
David: All right. So when you do explain this to someone at the proverbial party, are they generally satisfied with the— the notion that—
Luis: Yeah, I think most people, I think most people realize that it’s like, you know, these things are kind of like a— like a key. Nobody, nobody likes—it’s not like I love opening my door with the key. It’s kind of annoying. But it’s there, and I understand it just makes, it makes my house safer. In this case, it kind of just makes the whole Internet safer. eh–kinda gotta do it.
Thanks for listening to this second-to-last episode of the season. If you have any interest in a second season, please spread the word! Subscribe to this podcast. Leave a review on Apple podcasts, or a rating on Spotify.
UNSUNG SCIENCE with David Pogue is presented by Simon & Schuster and CBS News, and produced by PRX Productions. The Executive Producers for Simon & Schuster are Richard Rhorer and Chris Lynch.
The PRX production team is Jocelyn Gonzales, Morgan Flannery, Pedro Rafael Rosado and Ian Fox, project manager. Jesi Nelson composed the Unsung Science theme music, and Kristina Rebelo fact-checked my script.
At unsungscience.com, you can listen to every episode ever made, and read complete transcripts. For more of my stuff, visit davidpogue.com or follow me on Twitter, @pogue. Thanks for listening!
Podcast: Play in new window | Download
