At Amazon’s high-tech convenience store, Amazon Go, customers scan their phones at the door, grab the items they want, and walk right out. Their account is automatically charged.
Amazon’s “Just Walk Out” technology, powered by computer vision, sensors and deep learning, is a teaser for the future of physical retail, and is an early example of this year’s Technology Vision “Internet of Thinking” trend. The Internet of Thinking is about powering intelligent environments through edge and cloud computing, and Amazon Go offers us insight into how this will work in busy, real-world scenarios.
Inside the store, weight sensors on shelves monitor the movement of items. Dozens of cameras, custom-made to do computer vision work like motion detection and object detection, monitor activity from above. Between the sensors, the cameras, and the app, which shoppers scan at the doorway, there’s a lot of data being collected. Edge computing is key to filtering out unnecessary data and making the synchronization process more efficient.
But what happens if Amazon Go gets things wrong? When a CNBC journalist visited the store, she accidentally shoplifted yogurt. She had scanned her phone upon entry, picked up the yogurt, and walked out, just as instructed. But the store’s technology didn’t realize it, and so didn’t charge her. So far, Amazon has assumed responsibility for hiccups in their system, and told the journalist not to worry about it.
But small purchase discrepancies like this may be just the beginning of cashier-less technology’s challenges. As Amazon Go expands, the company will collect a wealth of data about customers. They’ll not only know what each customer purchases, but also their shopping habits inside stores, and even what they stop to look at and pick up, but ultimately don’t buy. By collecting this data, Amazon will have even more access to people’s lives than they do now. This increasingly personal view into their customers’ activity will increase the need for AI-driven customer insight systems to be responsible and explainable.
In environments where shopping—or any other activity—is monitored and managed by AI, it’s important to establish where responsibility lies. This year’s Tech Vision trend “Citizen AI” discusses how AIs must be raised to be responsible and explainable, especially as they are further embedded into society. Through Amazon Go, Amazon will have an intimate view of customers’ needs and interests: Consider what conclusions might be drawn about a customer who repeatedly picks up and puts down unhealthy items—especially if, from their Amazon.com order history, the company knows the customer has also recently downloaded a dieting e-book. Should the company’s recommendation system promote the items the customer has been picking up, but not buying, in the store? Should it promote healthier items that will help the customer stick to the diet he seems to be on? How much should the system disclose about what it knows about our habits and interests, and how it’s using that information? As a company’s role in our day to day life grows, their responsibility must too, or they’ll risk losing consumer trust.
Amazon Go has one store open in Seattle, with six more planned. And of course, they aren’t the only company exploring cashier-less retail; many are entering this area, using different combinations of technologies. BingoBox is a startup opening cashier-free convenience stores across China. Quartz reports that BingoBox is working to phase out their RFID system and move toward image recognition for automatic purchases. In the U.S., Kroger and B.J.’s Wholesale Club have both implemented a Scan & Go system that allows shoppers to scan codes on items as they shop, and sometimes pay from their phones, skipping the checkout process entirely. And a U.S. startup called AiFi recently announced their own checkout-free solution for retailers. AiFi has an AI sensor and camera network-based system, and according to TechCrunch, could scale from all the way from a mom and pop shop to a big retailer.
Through these efforts, we see how the Internet of Thinking can build a frictionless and intelligent environment, and we also see the rising need for responsible AI. These trends will be critical to the success of such companies as they attempt to build intelligent environments.
To learn more about the Internet of Thinking, Citizen AI, and our other trends, check out the full 2018 Technology Vision at http://www.accenture.com/technologyvision.