CHAPTER 1Your Livelihood
Going Places
On a September morning, the kind of equatorial summer day where the air is thick with the threat of rain and your clothes stick to your skin by nine o’clock, Ian Koli is waiting for me outside Connie’s Coffee Corner, a busy cafe in the Kibera neighbourhood of Nairobi, Kenya. As I introduce myself, we are joined by Ian’s friend and former co-worker Benjamin Ngito, who is loping towards us with an arm outstretched in greeting.
Ian and Benja know each other from their time working together at Sama, a US non-profit that outsources digital work to East Africa. Benja, who wears a two-day-old beard and a faded Superdry T-shirt, tells me he started his journey with Sama right here where we are standing, by Connie’s. Back in 2008, Sama’s local recruitment team promised him a fee if he could sign up twenty young locals for IT training at an internet cafe next to Connie’s. He only managed to find nineteen volunteers. ‘So I signed up. I had no choice,’ he said. ‘I needed that cash.’ He ended up working at the company for five years.
Ten years later, when Ian heard about a potential job, he thought it might involve admin or cleaning, and was surprised when his friend said it was artificial intelligence. ‘I had no idea what AI meant,’ he says. He’d never had an office job, let alone one in technology, having spent his teenage years working a series of informal jobs – cleaner, bricklayer, grassroots political organizer. When he couldn’t find work, he took small change from local politicians to instigate neighbourhood ‘chaos’ around election time, things like barricading roads, burning tyres or throwing stones at police. ‘In the ghetto here, it’s hand-to-mouth, you get paid cash and put it in food,’ he says.
Ian knew, from talking to Benja, and other friends, that working for Sama had changed their lives. Maybe, he thought, it could change his too and he would be able to move out of the place he was sharing with six other young men, and perhaps even save for the future. So he signed up to work for Sama and has worked there ever since.
When we’d met virtually during the coronavirus lockdown two years ago to talk about Sama and his work, Ian had been a skinny kid with a shy smile and a scraggly moustache. Today, he is walking us to his new home which he shares with his wife and four-month-old baby in the heart of Kibera. ‘Things are different,’ he says.
The neighbourhood of Kibera in central Nairobi is an informal settlement, or a slum, one of the largest in Africa that houses some of the country’s poorest families. This undercity of a million people is constantly moving, a flowing stream of camaraderie, haggling and humanity. We bob along like paper boats on its currents. Ruts and ditches are paths to be used, not dodged. On these narrow paths, pedestrians must defer to boda-bodas, or motorcycle taxis, imperiously honking vans, and kids kicking footballs. Butcheries and barbershops vie for space with women’s hair salons and chicken shops. They all advertise M-PESA services, the African digital wallet that is ubiquitous here. A pungent scent of slick garbage, heat and humans sits heavily in the still air.
Kibera is a complex, amorphous organism with its own villages, tribes and social classes. There’s an unspoken hierarchy here. Up the hill from where we are, in Laini Saba, crime is rife and houses are made of mud or canvas sheets, six or seven people to a single dwelling. But down here in the mini-village of Gatwikira, you can have your own shack that you share with maybe one or two others, made of sheet iron or eventually brick walls, and you can walk the streets in daylight without fear.
Kiberans live to survive, Ian tells me, fighting over the meagre amounts of water, electricity and jobs they are forced to share. But what unites them is a fierce loyalty to their neighbours and a collective mistrust of the state. Disputes are settled by local leaders, known as elders. They call the politician who’s been in charge here for twenty years ‘Baba’, Father.
I’ve never been to Nairobi before, but I grew up in Mumbai. And somehow this place – its entrepreneurialism, its everyday sorrows and big-hearted joys – reminds me of the home I’ve left behind.
As we arrive at Ian’s place, he jerks his thumb upwards at some wooden stairs. ‘First floor,’ he smiles. The elevation, a rarity in these parts, is a matter of pride. We ascend and duck through a narrow makeshift tunnel flanked by shacks, a boulevard of broken roofs. It smells of fresh soap. The residents outside, all women, are focused on the task of hanging up piles of dripping, clean washing. Some have infants tied to their backs. They nod in greeting.
Ian leads me to the last home on the left. A single naked bulb lights the neat space. Kibera’s soundtrack of hip-hop is muffled and faraway in here. A whirring desk fan roars in the sudden quiet. ‘My home,’ Ian says. ‘Karibu sana.’ You’re very welcome.
Every square inch of Ian’s home is perfectly utilized. The room comfortably fits a couch, two chairs, an upturned wooden crate that acts as a desk, and a large bed in the corner, curtained in patterned paisley for privacy. Along one wall, a dozen pairs of Ian’s trainers are displayed in pigeon-hole shelving, with baseball caps hung neatly underneath. At the foot of the bed, hidden from view, is a stove where the family cooks their meals.
The laptop sits on an elevated shelf next to the large TV, a deity of sorts. Netflix is loaded up, mutely cycling through ads for a series of Hollywood and Bollywood films and series. In late 2020, Sama teamed up with local telecoms providers to lay fibre broadband in large swathes of Kibera and elsewhere in the city to allow its agents to work from their homes during the pandemic. Ian was one of the workers whose homes went online, so he suddenly became popular amongst his neighbours.
‘This is my mobile office,’ he says. ‘I wake up here, work through the day, finish up, then I have time to go to school. I want to learn to code.’ Last year, Ian won a scholarship from Sama to attend college, where he is now studying for a bachelor’s degree in IT.
Ian’s job at Sama is to perform data annotation: he helps to train artificial intelligence software made by global corporations, by creating detailed labels for the datasets used to train them.
Ian works primarily on image-tagging for driverless cars. The computers inside these cars, developed by the likes of Volkswagen, BMW, Tesla, Google, Uber and others, need to know how to read a road – street signs, pedestrians, trees, road markings and traffic lights – so they can control the car’s driving functions. Ian usually receives driver’s-view clips of cars driving down anonymous roads, a bit like a hazard perception test for learners. The accompanying instructions ask him to tag every single object he can see by drawing bounding boxes around them; he draws little rectangles around all visible objects in the footage – vehicles, people, animals, trees, street lights, zebra crossings, bins, houses, even the sky and clouds.
The tasks remind me of the endless hours of ‘I Spy’ I play with my toddlers while on the move, with their little voices calling out triumphantly, ‘fence’, ‘gate’, ‘girl’, ‘puppy’, ‘truck’, as we drive or walk or ride. An hour of video might take Ian eight solid hours to annotate with labels.
While the work can seem repetitive, mindless even, Ian doesn’t mind it. ‘I found it interesting because I learned a lot about traffic rules and signs,’ he told me, knowledge he tucked away for the day he could drive a car. He’d also labelled the inside of homes, and various joints of the human skeleton. He didn’t need to know the names of the joints, he explained – just to mark them visually on an image.
As a kid, Ian loved playing with wires and electronics and had dreamt of being an electrical engineer. When he left high school, he’d had to support his mother and sisters and hadn’t had the money to go to college. ‘Now I wanna be a developer. When I joined Sama, I was imagining that these things that we are doing here, it is a stepping stone to Tesla, the company, or the Tesla technology itself.’
The mention of Tesla tickles Benja. ‘I saw that guy, Elon Musk, on TV. I said hey, that guy, I’m building his car!’
Ian wants to eventually start his own business, a dream of many Nairobi locals I meet both within and outside Kibera. ‘You know, causing chaos on the streets, that was the order of the day,’ he says. ‘You pick it up from your brothers and sisters. But after you have something that keeps you busy, your mindset changes, you cease to think like an ordinary ghetto boy, like an ordinary Kiberan, you think outside the box.’
Benja, too, was a rabble-rouser for hire, paid by local politicians to throw stones at the police. Now, he has just started his own walking tour company. But he also runs a chicken shop, leads a youth political activist organization and is opening a bar. He’s dabbled in selling water and electricity, lucrative businesses controlled by powerful cartels in Kibera. In his spare time, he’s a youth leader in Lindi, one of Kibera’s villages, where he helps prepare kids for formal office-based jobs. ‘I have been able to instil the Sama spirit and the culture into people that I meet in my area. And it goes on and on.’
Ian says he’s brought friends into formal employment too. ‘One guy, my school friend, was into pickpocketing, each and every day he was involved in these crimes,’ he says. ‘When he got the link from me, he joined Sama, he reformed drastically. If I tell you this was the guy, you won’t imagine it, you’ll say I’m lying.’
The mention of crime chastens Benja. He looks at Ian. ‘I wanna move out of Kibera by next year. And I want you to move too.’
‘It is in my plan, in three years coming, I should be out of Kibera,’ Ian says.
‘But then you have to speed it up.’ Benja cautions him. ‘You have a responsibility to leave. I always told you, hey, you don’t even need to be a team leader at Sama. You can be anything you want in this world. Man, I know you’re going places.’
* * *
As you drive down Mombasa Road, you trace the curves of Nairobi National Park, a wild oasis which makes Nairobi one of the only urban centres where you can spot giraffes alongside tall buildings while speeding down the expressway. Sama’s primary facility is on this road, four floors of a building in a large commercial business park, housing more than 2,800 people. Outside, a towering sign announces Sama, with its tagline ‘The Soul of AI’.
This building is all polished concrete floors and walls accented with corrugated iron. It is furnished in reclaimed wood and tin, with colourful hanging works of local art and potted plants. I’m told it’s supposed to be reminiscent of the workers’ homes in the informal settlements they come from. The designer, who consulted with early employees, wanted to use these familiar materials so that employees would think of the space as beautiful, and also as theirs.
Sama’s building is aesthetic, but it is ultimately an office. Banks of agents, the name Sama gives to its workers, sit at computers, clicking and tracing shapes around images of all kinds. Room after room is filled with twenty-somethings, young women and men, clicking, drawing, tapping. It requires precision and focus, but is repetitive, a game of shape-sorting, word-labelling and button-clicking. For human beings, these tasks are largely easy, obvious even, although for AI systems they are novel and complex. The agents confer sometimes, but mostly focus on their own screens, a few seconds per image and onto the next. Hip-hop streams out of one corner. The mouse-clicks drum along with the beats. A team of agents is tagging cars driving on streets in China and Japan; others are tagging close-ups of maize plants, satellite imagery of European towns, logging trucks lifting wood, and women’s clothing. Click, draw, tap.
An average work shift here starts around seven o’clock in the morning, lasting eight hours. Workers join Sama mostly from informal jobs like domestic cleaning, or selling chapatis on the street. Because of how the AI supply chain is broken down into bite-sized chunks, many of these workers have little, if any, visibility of the shape or commercial value of the final product they are helping to build. But they do know they are helping train software for some of the most advanced technological applications in navigation, social media, e-commerce and augmented reality.
For OpenAI, the creator of ChatGPT, Sama’s workers were hired to categorize and label tens of thousands of toxic and graphic text snippets – including descriptions of child sexual abuse, murder, suicide and incest. Their work helped ChatGPT to recognize, block and filter questions of this nature.
The agents work in teams of around twenty, annotating data almost continuously through the day, bar two scheduled breaks for food and drink. Outside of these, they are allowed toilet breaks but are otherwise expected to be at their desks. Team leaders are more mobile, milling about between rows, and looking over shoulders. At the end of each line, quality-control analysts spot-check the agents’ annotation work.
When it’s time for their scheduled lunch break, the agents troop noisily downstairs to the cafeteria hall, past signs saying ‘Silence please!’ and join a snaking line for their food. Today, there’s beef stew, with coriander rice, shredded cabbage in soy sauce, and mukimo, a Kenyan dish of mashed potato studded with greens. Slices of watermelon sweat in paper bowls. Everyone eats together.
I sit down to eat at a long table filled with chattering employees, including agents, team leaders and operations managers. Liliosa, a manager in her late thirties who assesses the company’s impact on its agents’ lives, is chatting about colonialism, the British royal family and Kenyan elections. She’s writing a hip-hop musical about a Kenyan freedom fighter who rebelled against the British. ‘Politics is our culture. And it’s tribal, each tribe wants their own to lead,’ she tells me. ‘But the youth don’t care anymore, they just want internet and jobs and money.’
After lunch, the cafeteria empties out rapidly, and I head back to the labelling floor. A young man is tapping through dozens of images of buildings around the world, Chinese pagodas and French apartments, marking whether they are historical or modern. For each image, he has to also click a series of boxes to describe the picture: moody, saturated, sharp or sepia. Click, click, click. He’s currently hovering over an image of an ancient Japanese Buddhist temple in Tokyo, standing behind a telegraph tower. It’s both, he decides, clicking on the option, a blend of history and modernity.
Each tap and click, I later discover, helps train algorithms that classify images for Material Bank, a platform for searching and ordering samples of architectural and design materials. The goal is to create an objective tool to pull out the most relevant information. It means when you search for a specific construction material or architectural style, the algorithm can serve you the perfect selection of examples you’ll need.
How does he know if he’s done it right? ‘Sometimes, it’s not clear,’ he tells me. ‘Then you just have to go with how you feel.’
The Ghost in the Machine
The pursuit of building intelligent, superhuman machines is nothing new. One Jewish folktale from the early 1900s describes the creation of a golem, an inanimate humanoid, imbued with life by Rabbi Loew in Prague, to protect the local Jews from anti-Semitic attacks.
The story’s consequences are predictable: the golem runs amok and is ultimately undone by its creator. This tale is resonant of Mary Shelley’s Frankenstein, the modern-day tale that helped birth the science-fiction genre, and of the AI discourse in recent news cycles, which is growing ever more preoccupied with the dangers of rogue AI.
Today, real-world AI is less autonomous and more an assistive technology. Since about 2009, a boom in technical advancements has been fuelled by the voluminous data generated from our intensive use of connected devices and the internet, as well as the growing power of silicon chips. In particular, this has led to the rise of a subtype of AI known as machine learning, and its descendent deep learning, methods of teaching computer software to spot statistical correlations in enormous pools of data – be they words, images, code or numbers.
One way to spot patterns is to show AI models millions of labelled examples. This method requires humans to painstakingly label all this data so they can be analysed by computers. Without them, the algorithms that underpin self-driving cars or facial recognition remain blind. They cannot learn patterns.
The algorithms built in this way now augment or stand in for human judgement in areas as varied as medicine, criminal justice, social welfare and mortgage and loan decisions. Generative AI, the latest iteration of AI software, can create words, code and images. This has transformed them into creative assistants, helping teachers, financial advisers, lawyers, artists and programmers to co-create original works.
To build AI, Silicon Valley’s most illustrious companies are fighting over the limited talent of computer scientists in their backyard, paying hundreds of thousands of dollars to a newly minted Ph.D. But to train and deploy them using real-world data, these same companies have turned to the likes of Sama, and their veritable armies of low-wage workers with basic digital literacy, but no stable employment.
Sama isn’t the only service of its kind globally. Start-ups such as Scale AI, Appen, Hive Micro, iMerit and Mighty AI (now owned by Uber), and more traditional IT companies such as Accenture and Wipro are all part of this growing industry estimated to be worth $17bn by 2030.1
Because of the sheer volume of data that AI companies need to be labelled, most start-ups outsource their services to lower-income countries where hundreds of workers like Ian and Benja are paid to sift and interpret data that trains AI systems.
Displaced Syrian doctors train medical software that helps diagnose prostate cancer in Britain. Out-of-work college graduates in recession-hit Venezuela categorize fashion products for e-commerce sites.2 Impoverished women in Kolkata’s Metiabruz, a poor Muslim neighbourhood, have labelled voice clips for Amazon’s Echo speaker.3 Their work couches a badly kept secret about so-called artificial intelligence systems – that the technology does not ‘learn’ independently, and it needs humans, millions of them, to power it. Data workers are the invaluable human links in the global AI supply chain.
This workforce is largely fragmented, and made up of the most precarious workers in society: disadvantaged youth, women with dependents, minorities, migrants and refugees. The stated goal of AI companies and the outsourcers they work with is to include these communities in the digital revolution, giving them stable and ethical employment despite their precarity. Yet, as I came to discover, data workers are as precarious as factory workers, their labour is largely ghost work and they remain an undervalued bedrock of the AI industry.4
Copyright © 2024 by Madhumita Murgia
