I've been helping to decode old books.. and didn't know it.. You probably are too..

[Fumesucker]

In a nutshell, some of those blurred out words you have to enter to post on some websites are actually scans from old books that are digitized but the computer can't decipher, humans are still better at reading funky text than computers are evidently.

http://news.bbc.co.uk/2/hi/technology/7023627.stm

A weapon used to fight spammers is now helping university researchers preserve old books and manuscripts. Many websites use an automated test to tell computers and humans apart when signing up to an account or logging in.

The test consists of typing in a few random letters in an image and is designed to fight spammers.

Carnegie Mellon is using this test to help decipher words in books that machines cannot read by letting sites use them to authenticate log-ins.

The test, known as a CAPTCHA (Completely Automated Turing Test To Tell Computers and Humans Apart), was originally designed at Carnegie Mellon to help to keep out automated programs known as "bots."

Read the rest of the article at the link.

[bananas]

1. Whoa!

What a great idea!
Thanks for letting us know!

[Frank Cannon]

2. Wait a minute. If the computer doesn't know what the word is...

then how can someone's input into the CAPTCHA be verified? :shrug:

[Fumesucker]

4. They show the same word to more than one person..

If two random people get the same answer then it's almost certainly correct..

If the two people do not agree they then go on and show it to a lot more people, I assume they then use a statistical process..

[joshcryer]

7. Here's the paper (it has grone exponentially from when it started):

http://www.cs.cmu.edu/~biglou/reCAPTCHA_Science.pdf

They do use statistics.

More history on Wikipedia: http://en.wikipedia.org/wiki/ReCAPTCHA

[Leopolds Ghost]

17. FASCINATING! In theory, this could have profound implications --

The wisdom of crowds could be used in various areas of cryptography and it sounds like this principle could be extended to cover all CAPTCHA's by using crowd wisdom to determine the correct answer.

[joshcryer]

5. It's tested against other users' input, not the computer.

[begin_within]

9. I've deliberately misspelled the words a few times and still been let through!

[Frank Cannon]

15. That's very interesting. I never tried that.

From now on, I'm only putting in obscenities.

[joshcryer]

3. This is from 2007. reCAPTCHA is now owned by Google, and does 200+ million CAPTCHA's a day.

http://en.wikipedia.org/wiki/ReCAPTCHA

In due course all data will be digitized. I'm kinda glad it's Google that's doing it as opposed to other companies or groups.

[Fumesucker]

6. I wasn't aware of it, that was the first article that popped up when I Googled..

[joshcryer]

8. No worries, it's cool man, it's a pretty remarkable service.

Google uses similar methods for their translation services (though they have a larger pool of data to go by they basically compare translated texts to one another).

[Leopolds Ghost]

19. Don't be glad. Google is a monopoly and is becoming more of one with their browser etc.

They also own YouTube now.

[eilen]

10. They should just hire out of work nurses

we have been decoding physician handwriting for years. We are highly skilled.

[begin_within]

16. Hats off to you

My Mom's doctor seems incapable of writing a simple prescription without at least one error that causes the pharmacy to reject it.

[hunter]

18. !

:rofl:

[Celebration]

11. must be why some of those are so hard to read

On some sites, those things are so hard to read that I have to try them several times. On other sites, they are really easy. Or maybe it depends on which book we are decoding????

[JVS]

14. Sometimes they are gibberish.

There are a couple ways of doing things. Some have one word that needs to be decoded, and some have two words that need to be decoded.

In the two word ones, there is always the key word (they know it and reproducing it will get you in) and the unknown word. They figure that if you put in the key word correctly, you'll put in the unknown word. What I find interesting is that sometimes the unknown isn't even from a standard english font. For example, I've seen plenty of words written in the old German script. Since I know that, I go ahead and enter them in correctly. Other times I have seen parts of mathematical fomulae involving superscripts and subscripts or even worse Hebrew letters (even if I knew them, what am I to do, transliterate into the Latin alphabet?) about which I have no clue. I just type random letters and it lets me through as long as I have the key word correct.

In order to make it uncertain which is they key word and which is the word they're trying to find out, they do distort the image of the key word a bit.

[HysteryDiagnosis]

12. They could scan "nuclear" and gw would never ever be able to type it out correctly. n/t

[Fumesucker]

13. Eh, if you gave a million monkeys a million keyboards..

They would all sound more intelligent than Dubya.

[Leopolds Ghost]

20. One *chimp is enough. n/t

[FredStembottom]

21. I noticed about a year ago that I was getting real words.

not just random characters.

Words like:

exonerate
indubitably
fallacious

That was a change that made me wonder!