Who really uses Mechanical Turks?

In a recent technical meeting at work, we discussed the Turk, a scam to make a mechanical chess computer (cough) that always wins. The secret is, of course, to secretly insert a master chess-playing dwarf inside the mechanism (don’t ask me where to get one of them).

Amazon went and built a Mechanical Turk service that allows people to defer tasks to real actual people programmatically (buzzword: crowdsourcing). A really novel idea for non-realtime processing that needs the complexities of human interpretation - but what is is used for?

I saw this article last night, and realized that there really were tangible uses for this service. The article doesn’t say who does the work, but it sounds like a prime target for crowdsourcing:

http://www.techcrunch.com/2008/03/29/image-recognition-problem-finally-solved-lets-pay-people-to-tag-photos/

And now suddenly TagCow appears, which allows users to upload photos and have them tagged within a few minutes. The technology appears to be “magic,” meaning there’s no explanation of it.

How is this system getting such intelligent results? Its recognizing the difference between hills and mountains, and the difference between witches and Ozzy Ozbourne - even I get that mixed up sometimes. But if I could supply a mugshot set of photos with names, all my photos could have tags of all my friends and family I’d consider paying for the service. And thats saying something, if you know how frugal I am.

Update: looks like they are using the Amazon service as expected!

http://www.techcrunch.com/2008/04/02/need-a-job-make-120hour-tagging-photos/

Leave a Reply