7
CAPTCHA the Magic
No comments · Posted by Johnny Malloy in Human, Online Personas, Optical Character Recognition, Validation
I started a little miniseries last week talking about the types of online personas and yesterday I continued with why Google would want to know if you were human or not. I just want to finish the thought concerning online personalities here by cataloging a few of the potential ways that humans are separated from automated processes in the online world.
The beginnings of separating artificial intelligence from human intelligence starts in the late 40s with Alan Turing and the Turing Test. What was a somewhat fantastical notion then is commonplace today. Often when entering form data you are challenged to solve a small puzzle to prove you are human. In an extremely interesting twist, researchers at Carnegie Mellon developed reCAPTCHA which not only identifies humans but slyly uses them to translate scanned text. For example, a great deal of the New York Times is not in digital format. We can try to scan this text but because computers can’t always read text (or the print may be slightly garbled) some of it doesn’t get translated. These bits are placed alongside CAPTCHAs for human translation, a brilliant bit of crowdsourcing developed in part by Luis von Ahn.
The strange coincidence of simultaneously using computers to test humans and humans to test computers has led to some interesting modern applications of process filtering:
- Ad companies can use CAPTCHA for branding.
- Twitter users can force followers to validate through an email response mechanism.
- Of course, everyone loves kittehs.
- CAPTCHA has even led to entire communities of mock-worship and a new religion.
- Not everybody likes CAPTCHA, and there are alternatives.
- But spammers have attempted to harness cheap labor to circumvent Turing Tests.
- They’ve also harnessed simple motivation techniques.
- And CAPTCHA has shown flaws maintaining a consistent filter of automated processes.
Once again, we see that using machines to identify humans turns into an arms race. Spammers even fight back using the exact same conceptual approaches, like bypassing email spam filters using images.
I love this quote from an interview in Walrus Magazine with Luis von Ahn, that he had “unwittingly created a system that was frittering away, in ten-second increments, millions of hours of a most precious resource: human brain cycles”.
This is exactly what Google and other search engines need in order to find the best content on the web: the real value of human brain cycles. We’ve seen the flaws of humans identifying machines, and likewise machines identifying humans. Until automated processes become indistinguishable from humans, the solution may simply be to just let humans to identify humans through actual interaction, one human being at a time.
No tags
