Amazon has upgraded its Just Walk Out retail shopping tech with a new AI model that’s not only more powerful, but could allow for checkout-free shops in more places.
The new multimodal code brings the transformer models often used in generative AI to the retail world. Rather than simply looking for the straightforward sequence of you picking something up and putting it back, Amazon is studying simultaneous data from cameras and sensors across the store.
The switch not only makes Just Walk Out more accurate, according to Amazon, but helps it overcome some of the problems with the original approach. It’s better at handling obstructions like nearby shoppers and poor lighting, and can otherwise handle odd scenarios like two people reaching for the same item.
Stores also won’t have to worry about retraining the AI if something unusual happens. A self-training approach uses both a 3D map and visual product catalogs to accurately determine what you’re buying.
Privacy is still important, Amazon claimed. You still have to present your payment card or a mobile wallet app at the entry gate, but Just Walk Out isn’t collecting biometric info. It only knows how you interacted with the store before you left.
You’ll likely still have to hunt for a compatible store at the moment. There are 170 third-party shops combined in North America, Australia, and the UK. Amazon plans to more than double that number through the rest of 2024, however.
The new AI could play a key role in that expansion. As the updated Just Walk Out is less dependent on the layout and lighting of a store, it should work in more locations without running into trouble.
The move comes at an unusual moment. While Just Walk Out reportedly helped drive more business at the Whole Foods stores that used it, Amazon has been pulling the technology from its grocery stores, including Whole Foods and its self-branded Fresh locations. The company is upgrading right as it’s scaling back its own locations.
Source: techopedia.com
Comments