There are no human cashiers, only cameras and automation technologies to process transactions. On Monday, the first intrepid customers enter Amazon Go, an Internet of Things shopping experience complete with cameras, sensors and an app.
Customers simply scan their smartphone at the electronic gates in the foyer, pick up the shopping items they would like and walk back out through the turnstiles. Amazon’s flagship store is in its newly opened Seattle headquarters, where some human staff replenish shelves of popular items, while others check ID in the alcohol aisle and greet newcomers at the door.
Technologies have replaced human cashiers; shoppers are automatically billed via the credit card connected to their Amazon account, detected as they leave through the gates. Multiple people can shop on the same person’s account by scanning in the account holder’s phone again, and in-store cameras recognise that person as authorised.
A major goal is to reduce queuing times at cashiers by adding automation technology to in-person retail.
Items for sale consist mainly of grab-and-go lunch options such as yoghurt, meat sandwiches and salads. Weight sensors detect when items are taken or put back onto shelves. Products placed back on shelves are automatically deleted from the shopper’s virtual cart. Hundreds of cameras follow shoppers around the store, able to distinguish between different shoppers by body type, although the company insist there is no facial recognition used. During a test phase, children placing items on incorrect shelves proved an unanticipated challenge, a source told the Sun.
IoT: Connecting the Dots
How do you solve a problem like Internet of Things platform security?
AT&T, Intel, IBM target safer Smart Cities with LTE sensors
Amazon Go Vice President Gianna Puerini said in an interview that the innovative set-up functioned well throughout the four-year test phase.
“This technology didn’t exist,” Puerini said, “It was really advancing the state of the art of computer vision and machine learning.”
“If you look at these products, you can see they’re super similar,” said Puerini, indicating two extremely similar Starbucks drinks beside one another on a shelf. One had light cream and the other had regular, and Amazon’s ML technology could successfully differentiate. The predominantly online retailer is keeping tight-lipped over the mechanics of the tech, divulging only that the shop runs on computer vision and machine learning software.
Customers can also press “refund” on their app and receive money back for an incorrect or faulty item. Amazon said the store partly operates under an honour system, and it anticipates thieves will be in the minority. A shop-lifting attempt by a New York Times journalist was immediately tracked by Amazon’s systems and the item was added to his bill.
If the trial 1800-square-foot (167-square-meter) shop is a success, Amazon said it will roll out the tech more widely. However, the retailer also said it has “no plans” to implement the tech in recently acquired Whole Foods stores in the US.