“Alexa, what time is it?”
It’s a simple question that any person with a watch can answer with minimal effort. But when you ask an Amazon Echo the same question, a vast system powered by natural resources and human labor is activated to drum up the answer. As many of us reckon with Silicon Valley’s impact on the world and consider how it has upended life, work, and even democracy, we also must consider the infrastructure–and the tangible harm it can do–that usually remains hidden beneath these seemingly simple user experiences.
It’s an aspect of AI that is nearly impossible to comprehend, let alone visualize, but a new map created by the co-founder of the AI Now Institute at NYU and AI researcher Kate Crawford and data visualization specialist Vladan Joler attempts this dizzying task anyway.
Called Anatomy of an AI System, the map and the corresponding essay lay out the components of the Amazon Echo, from the human workers mining the rare earth materials that power its chips to the black box of Amazon Web Services to the submarine internet cables that pass information across oceans. When you ask Alexa for the time, all these hidden pieces spring to life–but we rarely consider the consequences of such seemingly innocent questions on the global economic order and on the Earth itself, partially because it’s so difficult to understand.
“The complexity of tracing one consumer AI product is astonishing,” Joler tells Fast Company via email. “Every map, no matter how complex or precise it is, represents a simplification or reduction of the complexity.”
Joler and Crawford met at a retreat put on by the Mozilla Foundation, and they began talking about what it would take to visualize the entire system that undergirds voice assistants, something that’s completely obscured by the simple, rounded industrial design of the Echo and its competitors.
“The mineral extraction, smelting, logistics, fiber optic cables, networking, AI training, energy, and e-waste . . . it’s an almost impossible task, requiring a mind-boggling scale,” Crawford says. “So we started by drawing multiple version on butcher’s paper, and it took dozens of sheets.”
From there, Joler and Crawford took a year to research every piece of the Echo’s supply chain, uncover the hidden human labor that most of us don’t think about when we query a voice assistant, and put it in historical, geological, and anthropological context.
It’s not just the miners: It’s also the humans operating the gigantic global shipping and manufacturing apparatus that brings each piece of the puzzle together, it’s the click-workers who label and sort vast data sets on which to train AI, and it’s you, the user, who is simultaneously acting as “a consumer, a resource, a worker, and a product,” as Crawford and Joler write in the essay. Through this lens, Echo’s complex processing becomes a story of human work and–more disturbingly–human exploitation. A child laborer in the mines of the Congo would need to work for 700,000 years without stopping to accumulate the kind of capital that Amazon CEO Jeff Bezos makes per day. “At every level contemporary technology is deeply rooted in and running on the exploitation of human bodies,” Crawford and Joler write in the essay.
Visually, the duo represents this kind of exploitation using the shape of a triangle, which signifies how value is extracted and produced to contribute to the bigger whole. “Every one of those triangles is able to tell us a different story,” Joler says.” Unfortunately in most of the cases those stories are about brutal working conditions, exploitation, and the destruction of natural resources.”
It’s fitting that it’s difficult to read the map and uncover its intricacies on a computer screen. Joler recommends that you start with the Earth in its lower left corner and trace the entire life cycle of a consumer device that way. After the raw materials formed over billions of years, its tragic to examine how they are mined, used for a few years, and then dumped in a mangled, unnatural form.
The ultimate goal here is transparency, even if the two write that “in many cases, transparency wouldn’t help much–without forms of real choice, and corporate accountability, mere transparency won’t shift the weight of the current power asymmetries.” That’s because although tech companies love to pay lip service to transparency, they continue to deliberately obscure the systems making technology work, whether through trade secrets or information that’s simply kept from the public. For instance, Amazon has released almost no information about Amazon Web Services’ environmental impact, and it’s the largest cloud computing provider on the planet.
But without better understanding of the myriad layers that go into making a device like the Echo, there’s no way to interrogate it or improve it, for the sake of our fellow humans and our planet. From that perspective, Anatomy of an AI is a worthwhile start.
“We want to give people a range of access points to see these systems with fresh eyes,” Joler says. “Then we will hopefully be able to better imagine scenarios for a more fair and sustainable future.”