Flash Flash Revolution - View Single Post

Relambrien · 12-15-2010, 09:15 PM

Sup. I was bored, so I figured I'd share some information I've got about the thing Internet users love to hate.

So, CAPTCHAs. The word stands for "Completely Automated Public Turing test to tell Computers and Humans Apart" (though really, that's just the inventor trying to make it an acronym when the word comes from "capture"). Those who know what they are probably hate them. Those who don't probably also do without realizing it. You ever notice that when you sign up for a website, or even make a post in some places, you have to look at a picture and type in the letters in the picture? All of those are derivatives of the CAPTCHA. Their purpose is to make sure that you're a human being--computers (if the CAPTCHA is done right) can't correctly parse characters in the image, so automated bots can't figure out what the letters are and won't be able to type a correct response.

That said and done, here's some background information on CAPTCHAs. They were invented by a guy named Luis von Ahn, one of my professors at Carnegie Mellon University (which is how I know all this, his "introduce yourself to the freshmen" speech was all about it). They're based on the concept of the "Turing test", a generic term for something that allows computers and humans to be told apart from each other. The rest proceeded as the above paragraph implies. However, more recently, von Ahn has developed a new kind of CAPTCHA, called reCAPTCHA, which is so much cooler and more awesome.

reCAPTCHA is a method of using everyone on the Internet to perform a massively difficult task, specifically book digitalization. Computers and software have improved to the point where computers can recognize many words when scanned out of books, and can store the text in a digital form. However, there still exist a lot of words, especially poorly-scanned ones, or really obscure ones, that computers can't recognize. This is where reCAPTCHA comes in. A reCAPTCHA gives you two words: one of them is a word the computer already knows, and the other is a scan of a word it can't figure out. It matches your input against the word it knows to see if you're human, and assumes that you're right about the other word, which it then adds to its database.

The reason this works is because the word which the computer can't recognize is, by definition, impossible for a bot to get past. It's only there because it can't be recognized by a bot or computer. Then your answer becomes what the reCAPTCHA looks for when someone else gets that word. It examines the different results people put in and selects the one most likely to be correct, and uses that, so it's kind of hard to intentionally mislead. In this way, every time you fill out a reCAPTCHA, you're helping digitize some books, so be happy about it! As a side story, Professor von Ahn also told us that he received a series of emails from an illegitimate company in China that used bots to advertise itself, complaining that it had to hire a bunch of human workers to do what it used to be able to do automatically, just because of reCAPTCHA. So basically, reCAPTCHA created some jobs in China. Awesome, isn't it?

But that's not all von Ahn has done. He has an entire website devoted to similar concepts, Games With A Purpose. The most interesting/well-known one is the ESP Game. In this, two people are matched anonymously and shown the same picture. Each player has to type in words that describe the picture, and when two players match, they get a point. The goal is to get as many points as possible. This is a massive help to image search engines--if two people look at a photo and separately identify a dog, then a dog is probably in the photo somewhere and the word "dog" will produce a hit for that image. Google licensed this technology from von Ahn under the name "Google Image Labeler".

So yeah, check out the website, and try not to hate on CAPTCHAs so much. Especially reCAPTCHA, with all the good it's doing.

12-15-2010, 09:15 PM	#1
Relambrien FFR Player Join Date: Dec 2006 Location: Pittsburgh, PA Age: 32 Posts: 1,644	CAPTCHA and Games With A Purpose Sup. I was bored, so I figured I'd share some information I've got about the thing Internet users love to hate. So, CAPTCHAs. The word stands for "Completely Automated Public Turing test to tell Computers and Humans Apart" (though really, that's just the inventor trying to make it an acronym when the word comes from "capture"). Those who know what they are probably hate them. Those who don't probably also do without realizing it. You ever notice that when you sign up for a website, or even make a post in some places, you have to look at a picture and type in the letters in the picture? All of those are derivatives of the CAPTCHA. Their purpose is to make sure that you're a human being--computers (if the CAPTCHA is done right) can't correctly parse characters in the image, so automated bots can't figure out what the letters are and won't be able to type a correct response. That said and done, here's some background information on CAPTCHAs. They were invented by a guy named Luis von Ahn, one of my professors at Carnegie Mellon University (which is how I know all this, his "introduce yourself to the freshmen" speech was all about it). They're based on the concept of the "Turing test", a generic term for something that allows computers and humans to be told apart from each other. The rest proceeded as the above paragraph implies. However, more recently, von Ahn has developed a new kind of CAPTCHA, called reCAPTCHA, which is so much cooler and more awesome. reCAPTCHA is a method of using everyone on the Internet to perform a massively difficult task, specifically book digitalization. Computers and software have improved to the point where computers can recognize many words when scanned out of books, and can store the text in a digital form. However, there still exist a lot of words, especially poorly-scanned ones, or really obscure ones, that computers can't recognize. This is where reCAPTCHA comes in. A reCAPTCHA gives you two words: one of them is a word the computer already knows, and the other is a scan of a word it can't figure out. It matches your input against the word it knows to see if you're human, and assumes that you're right about the other word, which it then adds to its database. The reason this works is because the word which the computer can't recognize is, by definition, impossible for a bot to get past. It's only there because it can't be recognized by a bot or computer. Then your answer becomes what the reCAPTCHA looks for when someone else gets that word. It examines the different results people put in and selects the one most likely to be correct, and uses that, so it's kind of hard to intentionally mislead. In this way, every time you fill out a reCAPTCHA, you're helping digitize some books, so be happy about it! As a side story, Professor von Ahn also told us that he received a series of emails from an illegitimate company in China that used bots to advertise itself, complaining that it had to hire a bunch of human workers to do what it used to be able to do automatically, just because of reCAPTCHA. So basically, reCAPTCHA created some jobs in China. Awesome, isn't it? But that's not all von Ahn has done. He has an entire website devoted to similar concepts, Games With A Purpose. The most interesting/well-known one is the ESP Game. In this, two people are matched anonymously and shown the same picture. Each player has to type in words that describe the picture, and when two players match, they get a point. The goal is to get as many points as possible. This is a massive help to image search engines--if two people look at a photo and separately identify a dog, then a dog is probably in the photo somewhere and the word "dog" will produce a hit for that image. Google licensed this technology from von Ahn under the name "Google Image Labeler". So yeah, check out the website, and try not to hate on CAPTCHAs so much. Especially reCAPTCHA, with all the good it's doing.