So, Amazon’s ‘AI-powered’ cashier-free shops use a lot of … humans. Here’s why that shouldn’t surprise you
In 2021, when Amazon launched its first “just walk out” grocery store in the UK in Ealing, west London, this newspaper reported on the cutting-edge technologies that Amazon said made it all possible: facial-recognition cameras, sensors on the shelves and, of course, “artificial intelligence”. The first customers queued outside, excited to experience the future. “I am an early adopter,” one of them said. “I can’t wait to see how this new technology works and I think it is going to be everywhere shortly.”
The promise of the “just walk out” stores was that customers would not need to queue for a cashier, scan their own items or even pause on the way out. They could simply take what they needed, walk out the door and the benevolent all-seeing eye of technology would seamlessly price their goods, charge their account and send them a receipt.
The reality was that peoplewere watching Amazon’s customers shop. More than a thousand of them, as reported by The Information, watching the cameras and labelling footage of shoppers. An employee who worked on the technology said that actual humans – albeit distant and invisible ones, based in India – reviewed about 70% of sales made in the “cashier-less” shops as of mid-2022 (Amazon responded that “the characterisation of the role and number of human reviewers is not accurate”). Now, Amazon is reportedly moving away from “just walk out” and rolling out “smart shopping carts” instead (AKA a scanner in your trolley – big whoop).
I can’t stress enough how little of a surprise this should be. First, the fake robot shtick is very, very old. It goes back at least to 1770, and the original “Mechanical Turk”, a chess-playing robot that wowed the courts of Europe for decades until being revealed that it was, in fact, a series of grandmasters hiding in a box. Recent updates include Facebook’s “smart assistant”, M, which claimed to be AI but referred any complex queries to people; and Cruise, the self-driving car company whose operations required remote workers to intervene every two-and-a-half to five miles.
All of these are, separately, quite funny stories. But collectively they paint a picture of a society, and a culture, utterly unequipped to register the violence that is being done to it, merely because historical process is draped in the ribbons of “technology”. This violence is enacted simultaneously on the high street and the global stage. What makes me angry about how often we keep falling for it is not merely that we should know better, but what the costs of doing so actually are.
The national minimum wage in the UK is £11.44. A small grocery store like the Amazon Fresh shops might have half a dozen staff. Assuming all of them
were on full wage (unlikely) and all of them were on the lowest wage (ie not managers), the average individual salary would be about £20k and the annual wage bill would be about £130k. When this work is outsourced via video cameras, it is passed to data labellers. Amazon’s remote data labellers might be paid one or two pounds an hour, if they are lucky. If you can replace half a dozen UK staff with half a dozen data labellers in India, Kenya or the Philippines, then the difference in the annual staff bill alone could be almost £100,000 a year.
Jeff Bezos is the second-wealthiest person in the world, worth about $205bn (£163bn). That money doesn’t come out of nowhere. It doesn’t drop out of a pier-end slot machine called, “I learned to code at Princeton and that’s why I’m better than you”. It is the result of deliberately hiding actual work – designing, making, sorting, packing, cooking, farming, delivering – behind little icons on your smartphone screen, in order to devalue it. It is the systematic use of the fake robot trick to lower the value of labour, until people are reportedly sleeping in tents at the factory gates, then banking the difference.
The size of Bezos’s rocket is very precisely determined by the difference in costs between paying a worker in Britain and a worker in India – including all the historically determined racist and colonialist inequality that calculation involves. But make no mistake – Bezos and his ilk will pay a robot even less, as soon as that’s possible. The only lesson of Amazon Fresh is that we are not – quite – there yet.
The fake robot shtick has another purpose too: it’s a distraction. In 2021, Amazon and Google jointly signed a $1.2 billion contract to provide the Israeli state, including the military, with cloud computing and AI systems. While there’s no evidence that Google or
Amazon’s technology has been used in killings of civilians, this continuing deal displays a willingness to engage with a military that has killed 30,000 people, and whose use of “AI”-powered targeting allows it to say: “The machine did it.”
“Just walk out” might have had its day, but the elision of consumer comfort and plausible deniability is alive and well on the high street. Tesco opened its first GetGo store in 2021, promising the same kind of checkoutfree convenience as “just walk out” – they now operate stores in London, Birmingham and Welwyn Garden City. Tesco boasts that instead of using facial recognition, GetGo creates “skeleton outlines” of you. The underlying technology for the service is provided by Trigo, an Israeli company that boasts that almost all of its engineers were “cherry-picked from elite military units” including Unit 8200, the IDF’s military surveillance agency, and Unit 9900, its specialist surveillance and mapping division. Point-to-point tracking of unknown bodies through built-up space, based on algorithmic analysis of gait and posture? I wonder where they learned to do that.
James Bridle is a writer and artist, and the author of Ways of Being: Beyond Human Intelligence