167

Having worked at the Amazon Mechanical Turk for a long time, I find that some ReCaptcha questions are very similar to the 1c tasks there. Identify this, identify that.

A search shows that several web authors share the same suspicion. Most of them are high in the tinfoil scale, so they are not that reliable of a source. Included below for notability purposes.

claim source 1

claim source 2

claim source 3

WP/ReCaptcha#criticism

claim source 5

Given that Google is not a charity, and that recaptcha is free, the claim might hold water.

What is Google getting out of ReCaptcha? Free labour? Big data behavioural analysis? Something else?

  • 8
    They also use it to identify business addresses ("click images that have storefronts"), house numbers, and waterways ("click images where there are rivers") for mapping – user33109 Jun 1 '16 at 14:13
  • 45
    this is the entire point of recaptcha. – ell Jun 1 '16 at 16:35
  • 41
    This is amazing. You went on conspiracy theory blogs to see whether it was true, even though google makes no secret of this and it's explicitly written on recaptcha's website... – user31438 Jun 3 '16 at 10:07
  • 4
    "High in the tinfoil scale", that's rich. – Celeritas Jun 5 '16 at 10:06
  • 13
    The irony of this question is extreme. "Is Stack Exchange using Q&A sites as a free source of human-intelligent labour?" – Paul Draper Jun 5 '16 at 18:42
266

Yes, and ReCaptcha have always been open about it, before and after being acquired by Google.

From its formation, one of ReCaptcha's main selling points was that the data would be used. At first, it was used for fixing errors and ambiguities in the digitisation of books. Here's an example of this being praised back in 2007, 2 years before Google acquired it, when ReCaptcha was new:

reCaptcha makes captchas more useful than just preventing spam; by tapping into the reportedly 150.000 hours spent daily typing in captchas, reCaptcha has users proofread book text that OCR could not recognize and which would otherwise have to farmed out to a Mechanical Turk or other distributed proofreader. ...reCaptcha’s ingeniousness lies in making an otherwise cumbersome task worthwhile

Today (2016), it's expanded to include improving maps, machine learning/AI, and possibly other uses. Google are open about this, and even have a gallery of examples of how recaptcha data is used: https://www.google.com/recaptcha/intro/index.html#creation-of-value

Millions of CAPTCHAs are solved by people every day. reCAPTCHA makes positive use of this human effort by channeling the time spent solving CAPTCHAs into digitizing text, annotating images, and building machine learning datasets. This in turn helps preserve books, improve maps, and solve hard AI problems.

Google were also open about this back when they first acquired it. In fact, at the time, not only did their slogan reference how the data was used, but the tool itself also explicitly stated it:

reCAPTCHA. Stop spam, read books.

The words above come from scanned books. By typing them, you help to digitize old texts.

enter image description here

Screenshot from this blog

They promoted the product as a form of "crowdsourcing", using labour people were giving freely anyway on existing captcha systems to do something useful. For example, from the official blog post by Google from 2009 announcing the acquisition:

Since computers have trouble reading squiggly words like these, CAPTCHAs are designed to allow humans in but prevent malicious programs from scalping tickets or obtain millions of email accounts for spamming. But there’s a twist — the words in many of the CAPTCHAs provided by reCAPTCHA come from scanned archival newspapers and old books. Computers find it hard to recognize these words because the ink and paper have degraded over time, but by typing them in as a CAPTCHA, crowds teach computers to read the scanned text.

In this way, reCAPTCHA’s unique technology improves the process that converts scanned images into plain text, known as Optical Character Recognition (OCR). This technology also powers large scale text scanning projects like Google Books and Google News Archive Search.


Whether or not that's efficiency or exploitation, smart or immoral, is subjective - but there's never been any doubt or secrecy about the fact it happens.

[edit] of course, with some of the challenges coming from "text" software had failed to identify, there's always the possibility it might be unreasonably difficult or not even text at all. With thanks to Mateo's comment:

??

  • 18
    If this helps make a better OCR software, I'm all for it. I've done a lot of OCR result proofing and the current apps leave a lot to be desired. – Joe L. May 30 '16 at 15:54
  • 29
    What I don't understand is, doesn't it have to know what the answer is to verify your input? How then, are you helping by participating? – Carcigenicate May 30 '16 at 20:30
  • 36
    @Carcigenicate Follow the link below the image, and the comments: at least one person tried to sneak rude words into Google Books by figuring out which word was the "test" word and which was the "scan" word, writing the "test" one correctly to pass the test, while trolling the "scan" one. Presumably the devs got wise to this and only accepted input that matched from multiple unrelated people... – user568458 May 30 '16 at 20:39
  • 36
    @Carcigenicate as I understand it that is why they sent two words, one was a word they already had high confidence about and could therefore use as a test. The other was a word they wanted you to read but you didn't know which was which. IIRC they also sent "suspiscious" users two "test" words rather than one "test" and one "read". – Peter Green May 31 '16 at 1:02
  • 10
    The “unknown word” if entered the same way by a few people, then becomes a test word. As this is a word that the best OCR software could not cope with, the system generated its own test images. Creatiing the test images for CAPTCHA is one of the hard problems to solve. – Ian Ringrose May 31 '16 at 17:12
40

Yes!

Luis von Ahn, one of original developers, talked in one TEDx conference about reCAPTCHA technology, and his new Project DuoLingo

In this presentation, he talks about CAPTCHA history and problems and how people were wasting about 500,000 hours every day using CAPTCHA. Then he thought how use this time in a useful thing, like helping OCR books.

He wants use this idea about massive collaboration on Duolingo, a Language-Learning Platform. The idea is to translate texts to more languages than English.

You must log in to answer this question.

protected by Community Jun 6 '16 at 2:44

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged .