Amazon’s Just Walk Out tech coming to campuses in Australia

The Canberra Institute of Technology is implementing a ‘cashierless’ system in its Bruce campus cafe, raising the question; who exactly is Bruce?

Canberra Institute of Technology students with the late-night munchies will be pleased with the roll-out of Amazon’s ‘Just Walk Out' technology on campus, such that they will soon be able to pick up a Snickers at their student cafe at any hour of the day without the need for human interaction or potential concerns over self-checkout dramas.

The paranoia may however flare up if they ponder for too long the advanced AI facilitating their ‘frictionless’ transaction.

To be implemented by global consultancy Cognizant and claimed to be the first at an education institute in the Southern Hemisphere, ‘Just Walk Out’ leverages machine learning and computer vision technology to track the movement of individual shoppers and their selection of items through a system of cameras and sensors, assigning a one-time code linked to a customer’s payment device so that they can simply leave the store without removing their wallet.

However, should they take off their beanie for example while still wandering the aisles, no problem, the system updates their signature based on their other defining pixels. The technology can also track groups of shoppers sharing a single card, distinguish between different Shapes flavours through object recognition, and where an item might be too small to accurately identify, such as with chewies, it fuses vision with shelving weight-sensor data to get a positive read.

In addition to the obvious saving on staff wages, the system also provides a range of other benefits for proprietors. ‘Just Walk Out’ – no longer the old-fashioned philosophy of perennially cash-strapped students – helps to address light-fingered shrinkage, as well as streamline stock management. And in the event of an obnoxious customer dumping their suddenly unwanted Bega slices on a random unrefrigerated shelf, the store will be sent an alert to hopefully cut down on spoilage.

Amazon is quick to note that Just Walk Out doesn’t collect any biometrics, although behavioral characteristics are captured to provide analytics for example on how often a specific item is returned to the shelf, generating insights which can inform store planning and inventory decision-making. The system can also operate independently of the tech giant’s Amazon One hand-wave payment option, which uses a consumer’s distinct credit-linked palm print for identification.

“All we need to know is where a person is on the floor, and where their hands are in relation to the store’s merchandise,” assures Gérard Medioni, Amazon’s lead scientist on the design project. Kicking off in 2014, Medioni and his team created photorealistic synthetic ‘consumers’ and employed generative adversarial network AI technology to develop the system: “With GANs, you can control absolutely every movement that the virtual shopper is performing.”

The purpose was to train the tech on both realistic and potentially abnormal behaviour or occurrences such as sudden crowds, with the annual Boxing Day shit-fight springing to mind.

Now, Amazon says the system can recognise millions of actions without making mistakes, although its imagined a game might soon spring up on Australian campuses to see who can come up with the most elaborate scenario to confuse the robots, such as with the events depicted in the 2012 movie ‘Bait’.