When customers in the London borough of Hackney shop in the new Amazon Fresh store, they no longer pay a checkout operator but simply walk out with their goods. Amazon describes “just walk out shopping” as an effortless consumer experience. The rise of automated stores during the pandemic is just the tip of the iceberg. Floor-cleaning robots have been introduced in hospitals, supermarkets and schools. Fast-food restaurants are employing burger-grilling robots and chatbots. And delivery bots are being rolled out at an accelerated pace. As Anuja Sonalker, chief executive of Steer Tech, a tech company specialising in self-parking, ominously said last year: “Humans are biohazards, machines are not.”
With the realisation that machines are immune to viruses and social distancing, we have seen the return of an apocalyptic consensus: according to one recent prediction, as many as half of all work tasks are at risk of automation by 2025. Such gloomy forecasts conjure a world where robots do all the work and humans are consigned to history’s dustbin.
We’ve been here before. Throughout capitalist history, times of crisis have bred anxieties about robots stealing our jobs. After the 2008 financial crash, a series of studies pointed to an automation tsunami that would swallow as much as half of the world’s work in the coming decades. Although that much-prophesied dystopia has not yet arrived, a scenario less spectacular but equally grim is growing in its shadow: the rise of microwork. In short, microwork refers to the human “jobs” that involve nudging artificial intelligence in the right direction. Workers, mainly in the global south, sit at computers clicking on images that, for instance, show autonomous vehicles how to navigate city centres, facial recognition cameras how to spot emotions, and marketing software how to spot breeds of horse.
“For a penny, you might pay for a person to tell you if there is a human in a photo,” Jeff Bezos explained, at the public opening of Amazon Mechanical Turk (MTurk), the first and most infamous of such sites. Like other similar platforms, such as Clickworker, which match underemployed and jobless people with online piecework, Mechanical Turk operates on a simple premise. The platform hosts contractors, often large tech companies such as Twitter, who outsource short data tasks such as labelling images – lasting a few seconds to a few minutes – to workers with few labour rights or secure hours.
Such sites have seen a boom in users during the pandemic. At a time when many have lost their jobs and are stuck inside, work that only requires an internet connection and a laptop can offer a much-needed source of income. The platforms often present the work as the preserve of glamorous young freelancers. But hazy promises of the remote -work dream disguise a brutal reality. Many workers on these sites have few other options, or are otherwise excluded from the formal economy. They may reside in poor rural areas, prisons or refugee camps, and find microwork through non-governmental programmes that aim to “Give Work, not Aid”. A World Bank researcher in 2012 wrote of a situation where millions of tiny digital tasks generated “thousands of jobs”. But microwork is often so sporadic and poorly paid it can hardly be called a “job”. In 2018, formerly middle-class Venezuelans facing an increasingly desperate economic situation sat at laptops and annotated images of urban areas to train autonomous vehicles. Workers were paid by the task and, in some cases, made less than $30 a week.
In many respects, the work differs little from the survivalism of “wage hunters and gatherers”, who spend their days doing a dizzying range of odd jobs such as shoe-shining, selling tissues and picking litter. With jobs on microwork sites lasting as little as a few seconds, workers must continually hunt for work, and might be contracted by upwards of 50 “requesters” over the course of a day. Carved into tiny segments, the jobs are opaque, often surreal and sometimes humiliating. One task on Mechanical Turk allegedly asked workers – or “Turkers” – to post pictures of their feet for reasons unexplained.
Opacity, however, is no software glitch. By design, the platforms obscure operations and preclude worker organisation, promising contractors a dream scenario: all the work without the troubles associated with an actual workforce. Impenetrable rating systems, which permit contractors to reject “bad” tasks out of hand, only allow workers to contact and “challenge” the contractors, who are under no obligation to reply. Wage theft is thus all too common – a report by the International Labour Organization found that on one major platform, about 15% of all tasks go unpaid.
In a statement, which has been edited for length, Amazon Web Services said, “MTurk is a marketplace where requesters determine how much they are willing to pay a worker to complete a specific task. The amount of compensation workers receive depends on the price requesters set, the number of tasks workers complete, and the quality of their work. Most workers see MTurk as part-time work or a paid hobby, and they enjoy the flexibility to choose the tasks they want to work on and work as much or as little as they like. While the overall rate at which workers’ tasks are rejected is very low (less than 1%), they also have access to a number of metrics that can help them determine if they want to work on a task, including the requester’s historical record of accepting tasks.”
The freedoms many of us have enjoyed working from home during the pandemic are the flipside of new kinds of control and surveillance. Meetings on Teams and Zoom send data straight to Microsoft and Amazon. Militant bosses have made employees keep their webcams on to display their faces and keystrokes. Like the workers on microwork sites, our labour is increasingly captured as data to power artificial intelligence. How the data is then used remains a mystery. Maybe to directly show AI how to do our jobs; or perhaps to expose AI to data about the emotions we experience at work. One thing seems clear: increasingly the primary or secondary role of work is no longer just work, but to show robots how to do our jobs, even if this aspiration in many cases remains a far-flung fantasy.
But the picture is not wholly bleak. Just as these sites act as experimental labs for new forms of exploitation and control, they also generate new strategies of resistance. In lieu of union representation, workers resort to letter-writing campaigns to draw attention to their work, forums that challenge the platform, and browser plug-ins to spotlight unscrupulous contractors. Online forums become loose networks of support that offer advice and guidance to platform users. These tactics remain in their infancy. But as all our jobs are increasingly driven by the demands of “big data”, we will need similar tactics to wrestle back some control of our working lives – as well as taking notice of those who make our digital lives so seemingly “effortless”.