Amazon has developed a new AI model for its “Just Walk Out” technology that improves the accuracy and efficiency of its checkout-free approach to shopping. The new multimodal AI foundation model analyzes different types of sensor data simultaneously, allowing for the automatic identification of items that shoppers take from store shelves. This new system aims to streamline the process of tracking shoppers’ activity and generating receipts and is set to roll out to existing stores over the next month.
The new AI model is based on transformer-based machine learning models used by generative AI applications and has been in development for more than two years. Amazon demonstrated how the new AI model works during a briefing for reporters at its campus in Seattle. The company detailed the new system in a blog post and mentioned that it will be used in Amazon’s first-party retail locations, such as Amazon Go convenience stores, as well as in third-party locations, including airports, stadiums, universities, and hospitals.
Amazon’s Just Walk Out technology team operates independently from the physical stores team as part of the Amazon Web Services cloud division. The technology is known as JWO within Amazon and utilizes a combination of cameras and weight sensors in shelves, along with a catalog of store items. Just Walk Out is currently available in 170 third-party locations in the U.S., the UK, Australia, and Canada, and aims to double the number of third-party stores this year.
The system uses three-dimensional interactive maps of stores that can be generated with the built-in LiDAR in cameras on Apple devices. The technology has been tested in various scenarios, including situations where customers grab multiple items at once or when the view of an item is obstructed. Amazon has also developed an RFID-enabled version of the technology, currently available at several stadiums in the U.S. and internationally, with plans to expand to more venues in the coming months.
Just Walk Out aims to reduce lines, enable 24/7 operations, reduce theft, optimize labor, and create more efficient retail space in new venues for third-party retailers. Amazon has debunked reports that humans are watching through cameras to tally up each shopper’s bill, clarifying that people are only used for annotations in the training process for the AI. The system itself operates autonomously and uses a combination of cameras, weight sensors, and a catalog of items to track shoppers and generate receipts accurately and efficiently.
Overall, Amazon’s new AI foundation model for its Just Walk Out technology is set to improve the accuracy and efficiency of its checkout-free approach to shopping. With the new multimodal AI system, Amazon aims to streamline the tracking of shoppers’ activity and generate receipts faster and more accurately. The technology is set to roll out to existing stores over the next month and is based on transformer-based machine learning models used by generative AI applications. Amazon is committed to expanding the use of Just Walk Out in its first-party and third-party retail locations, offering a more convenient and efficient shopping experience for customers.
