Amazon Proves It Remembers Mechanical Turk

If you have an iPhone and a few spare minutes, you should check out Amazon’s new app. Snap a photo in a store or on a friends’ book shelf and Amazon will find you the item for sale on its site. Gadget Lab has a full look at the feature. Behind the scenes, it’s not […]

If you have an iPhone and a few spare minutes, you should check out Amazon's new app. Snap a photo in a store or on a friends' book shelf and Amazon will find you the item for sale on its site. Gadget Lab has a full look at the feature.

Behind the scenes, it's not Amazon employees, or any artificial intelligence doing the legwork. It's ordinary folks around the world, working for pennies, using Amazon's Mechanical Turk.

While the service is several years old, this is perhaps the best use case yet for Turk. A quick browse of the more than 30,000 tasks currently available gives the impression that most use the service for the seedy underbelly of the web. You can earn a penny by rewriting a paragraph of text, or make three times that just for clicking on an ad.

My two tests of the Turk-powered Amazon Remembers feature on the iPhone had good results. In about five minutes it found a book on my desk and my exact brand of water bottle. I never saw either of my pictures show up on Turk, though tasks tend to flow quickly.

Before you become too enamored with Amazon's technology, another iPhone app found the book in about three seconds. SnapTell uses image recognition to immediately reply with the best match. Product packaging and book covers make this process easy. Expect Amazon to utilize a similar technology as a first pass, if it isn't already.

SnapTell was unable to find my water bottle, on the other hand, proving there still is a need for humans, for now.