Millions of jobs in technology don’t fit the typical image you may have in mind when you think of someone employed in the digital sector. Instead of degreed analysts and qualified coders with clout and privilege, there are many tech-enabling functions that are fulfilled by people, but are rendered largely unseen by the eclipsing power of words such as “innovation” and “artificial intelligence” (AI).
I’m not just talking about jobs in warehousing and fulfilment of e-commerce. There is much necessary focus on Amazon and Amazon-type warehouse workers. As I write this, Amazon workers in Alabama are in the final hours of voting on whether to unionise, a fight that’s been gearing up for years and involves shocking allegations of worker mistreatment — which, naturally, Amazon denies.
Still, there is an even more unseen group: workers who process the data that feeds into AI and other machine-learning systems; people who moderate content on social platforms; those who label and describe images for computer vision and image-processing algorithms; the real people who enable and check natural language processing. These are all functions that — for now — are still beyond the capabilities of tech alone.
Take computer vision, for example. Interpreting the content and meaning of an image has long been imagined as a possible processing role for computers. Yet perfecting it has been a far greater challenge than expected. There is complexity (both visual and psychological) in this task that humans do in the blink of an eye. As author Jason Brownlee explains in Machine Learning Mastery: “A given object may be seen from any orientation, in any lighting conditions, with any type of occlusion from other objects, and so on. A true vision system must be able to ‘see’ in any of an infinite number of scenes and still extract something meaningful.” This is seen as essential technology in achieving truly autonomous self-driving cars or augmented reality glasses.
Interpreting the content and meaning of an image has long been imagined as a possible processing role for computers. Yet perfecting it has been a far greater challenge than expected
It still often takes human judgment to categorise a product or decide if an image is offensive. Advanced computer vision would be able to recognise not just the subject of an image but also elusive and amorphous qualities, such as mood or gestures, the signifiers and connotations of which are further complicated by things such as culture and context. For now, we need people to do the work of processing a huge range and quantity of data, and feed that into digital systems until they can take over. This is the “microwork” category of digital work.
Whereas macrowork tends to be discrete (stand-alone) skilled work, such as programming or design, microwork is jobs that can be split up into small tasks and repetitive functions, often distributed to many people. Microwork is simple enough to be farmed out to a largely unskilled or non-technical workforce, with little training, but beyond the ability of automation.
Amazon has a finger in this pie too. Mechanical Turk (or MTurk) is an Amazon-owned “crowdsourcing marketplace” that connects employers to a distributed and virtual workforce. Similar companies include Crowdflower, Microworkers and Sama.
Almost a decade ago, the Harvard Business Review argued that “impact sourcing” (distributing microwork to the “bottom of the pyramid”) could be the answer to “break[ing] the cycle of poverty”. It hasn’t quite lived up to that yet. Rather, some argue that these practices could be deepening divides — Mary L Gray and Siddharth Suri co-authored Ghost Work, which warns about the creation of a “new global underclass”.
The Online Labour Index (OLI), which describes itself as the provider of “online gig economy equivalent of conventional labour market statistics” reveals that most online labour employers are based in the US, but the biggest supplier of online workers themselves is India (24% of workers observed) followed by Bangladesh (16%). India has long been associated with outsourcing.
Cheap labour
Furthermore, while software development is still the largest category of work on the OLI, it is clerical and data-entry work that has shown a sharp increase in recent months.
Affordability factors right in there, with employers in developed countries able to build their dominance and profit based on the grunt work farmed out to relatively cheap labour in developing countries such as our own. This is why a number of activists and organisations are calling for regulation of ghost work.
In February, Alexandrine Royer from the Montreal AI Ethics Institute wrote a piece for public policy non-profit organisation Brookings, in which she argues that this is creating “downward pressure on wages” and that these workers face “economic insecurity, unpredictable working conditions and limited bargaining power”. They are, she says chillingly, “the new extractive commodity”.
Even if you care nought for the plight of workers, the judgments these ghost workers make feed directly into AI ethics and inform the algorithms that will increasingly shape our world — another compelling reason to lift the veil on this hidden industry.
• Thompson Davy, a freelance journalist, is an impactAFRICA fellow and WanaData member.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.