Ghosts in the machines: the invisible human labour behind AI

The greatest risk may not be that computers will one day think like humans, but that humans will forget about the workers sustaining our digital economy.

Sign Up

Get the New Statesman's Morning Call email.

At the AI: More than Human exhibition at the Barbican Centre in London, I spent a while playing with Aibo, a glossy silver robotic dog manufactured by Sony in Japan. Aibo, which stands for artificial intelligence robot, can chase a ball, wag his tail and, if you’re lucky, put his paw in your hand. It’s hard to tell whether he is responding to humans. His wide-set LED eyes are inscrutable. Aibo may be cute, but he definitely doesn’t have a brain.

Computers that can actually think like humans – or pets – continue to elude scientists, yet the existence of artificial intelligence (AI) has become received wisdom. Utopian technology books suggest an ineluctable arc of progress between Aibo-like passivity and conscious computation. David Cameron, our former prime minister, recently announced that he will soon take up an advisory post at US-based AI company Afiniti, registering his excitement at AI’s “rapid development” and “huge potential” to address social challenges.

In their recent book Ghost Work, Mary L Gray and Siddharth Suri argue that AI is a long way from producing autonomous human intelligence, and indeed depends on significant human intervention. Beyond the Amazon Echo assistant in your kitchen or the disembodied apps that glow on your iPhone screen are a hidden army of “ghost workers” who intervene when algorithms trip up. This invisible labour force administers complicated takeaway orders, moderates explicit Facebook content and verifies your Uber driver’s picture when an algorithm fails to recognise their new haircut.

Who does this kind of work? People in the US, India and elsewhere, working from their bedroom or kitchen counter, connected to the internet and often earning less than the minimum wage. Ghost workers are pivotal to digital capitalism: the Pew Research Center estimates they number 20 million people.

“AI systems are technically illusions, because they’re always in a state of becoming rather than being,” says Fabian Ferrari, a researcher at the University of Oxford. Rather than replacing dull, low-paid human tasks, AI merely conceals them from view. In their book, Gray and Suri term this “the paradox of automation’s last mile”. When AI advances, it creates new temporary labour markets for human fixers.

Amazon’s Mechanical Turk, a crowdsourcing digital marketplace, is one of the emblems of ghost work. It allows employers, so-called requesters, to post “human intelligence” tasks such as tagging images, entering data and completing surveys. “You’ve heard of software as-a-service – well this is human-as-a-service,” Amazon’s chief executive, Jeff Bezos, boasted to an audience in 2006 when launching the platform. Its name stems from an 18th-century chess-playing machine that allegedly beat Napoleon Bonaparte. The machine was later revealed to be a trick: a human chess master would hide inside its cabinet, controlling the movements of a robotic dummy. Bezos’s modern-day version similarly fulfils the deferred dreams of computer intelligence with humans rather than machines.

It’s easier and cheaper to employ humans to behave like machines than it is to develop machines that simulate human behaviour. Of course, many technology companies would rather you didn’t know this. Venture capitalists invest in the idea of human obsolescence. To them, routine human labour is an ungainly truth – the future, after all, lies in intangible capitalism, where returns flow to platform owners unrestricted by organised labour.

In 2016, for example, Bloomberg documented the human workers who spend 12 hours a day pretending to be chatbots for a start-up that claimed its personal assistants were powered by AI. In 2017, the business expense management app Expensify admitted using humans, outsourced through Amazon Mechanical Turk, to transcribe receipts it claimed to process using “smartscan technology”.

Our wonder at AI reveals much about the way we value some types of work and erase others. Ghost workers do the mundane jobs that our economy was supposed to automate, undertaking the hidden digital labour that allows a new creative class of start-up founders and internet influencers to thrive. The greatest risk may not be that computers will one day think like humans, but that humans will forget about those of us performing the monotonous, enervating work that sustains our digital economy.

Hettie O'Brien is a New Statesman online editor. 

This article appears in the 07 June 2019 issue of the New Statesman, The Trump alliance