(Bloomberg Opinion) — You may have heard that revolutionary AI rests on old-fashioned foundations. The supply chain that produces generative AI tools like ChatGPT has highly paid executives and researchers at the top, and people at the bottom toiling to screen training algorithms. According to a recent World Bank estimate, between 150 and 430 million people do this kind of work: annotating images, text and audio; create frames around objects in images and, more recently, write haikus, essays and fictional stories to train the advanced tools that could eventually replace people like me.
They are also in something of an economic standstill. “I've never met an employee who said to me, 'This job gave me the chance to buy my house or send my kids to college,'” says Milagros Miceli, a researcher at the Distributed AI Research Institute and Weizenbaum Institute , who has worked with dozens of data workers around the world.
Miceli remembers speaking in a slum in Argentina in 2019 with a dozen data label workers who earned about $1.70 an hour. When she returned in 2021, no one had left and their wages had barely increased. They still lived below the poverty line.
Workers often have to take on second jobs or night shifts, says Madhumita Murgia, the AI editor of the Financial Times, whose recent book Code Dependent features their stories from across the developing world. For example, a woman working for Samasource Impact Sourcing in Nairobi was unable to support herself and her daughter on her salary and had to move in with her parents, Murgia says.
The job itself is precarious. Another worker in Bulgaria was unable to pay rent because she was suspended from taking paid duties after complaining about night shifts. “You're just one step away from everything unraveling,” says Murgia. End customers include Microsoft Corp. and OpenAI, some of the most valuable companies in the world. “It's like the factory worker in the Philippines who doesn't realize that the dress they are sewing is going to be a $3,000 dress.”
There is also very little of that age-old aspiration for developing countries: upward mobility. Murgia found that data workers weren't moving to higher-paying digital jobs. “They are still limited to low-quality work,” she says.
Leaders of data labeling companies often start with noble intentions to help people out of poverty, but they struggle to get business customers to pay higher rates as competition in their field has increased. As such, most data work platforms do not have policies in place to ensure their employees earn at least the local minimum wage, according to a 2021 study from the Oxford Internet Institute.
Take this job posting for “professional translators” in Igbo, Nigeria, which offers up to $17 per hour to help train generative AI models. That's well below the average rate for Nigerian translators, who start at $25 per hour, according to Good Firms, a customer review website. The ad comes from Remotasks, the flagship platform of San Francisco-based AI startup Scale.ai, which just raised $1 billion from investors including Amazon.com Inc. in one of the largest funding rounds of the year. Scale.ai did not respond to multiple requests for comment.
The company and rivals like San Francisco-based Samasource Impact Sourcing Inc., Argentina's Arbusta SRL and Bulgaria's Humans in the Loop play a crucial role in the AI supply chain, but for years have typically paid workers just enough to make a living to provide. Murgia and Doctor Miceli say.
This can remain the case even as data work becomes more complex. Recently, platforms like Scale.ai have started looking for more skilled workers, including artists and people with creative writing degrees, to write short stories for training AI systems, according to instructional documents seen by Miceli. While these offer higher wages, they are still below what people with a degree should earn.
Researchers say interest in such work is growing, but because there are few incentives to provide fair wages, it is difficult to see workers' economic status improve. Training AI is already terribly expensive due to the costs of chips and cloud computing. (Venture capital firm Sequoia Capital recently calculated that the AI industry spent $50 billion on Nvidia Corp. chips to train AI by 2023, but only generated about $3 billion in revenue.)
That means fewer opportunities for the people driving the AI revolution and shows once again that the true transformative effects of technology lie in entrenching economic power.
Maybe we can learn something from Nike Inc. In the 1990s, the company faced a huge backlash due to the long working hours and meager wages earned by workers in the developing world. Over time, consumer boycotts and media pressure led Nike to implement stricter labor policies. It spent millions of dollars improving conditions and wages.
The challenge for data workers is that their work is harder to visualize in the same, concrete way that you might imagine a young boy sewing tennis shoes in a dimly lit warehouse, and that can make it harder for their advocates to gain support. But tech companies should remember that poor working conditions at the bottom of their supply chain can also lead to substandard AI. That's problematic at a time when the public is more wary than ever of bubbly models who hallucinate. The answer to that is not rocket science: pay the data workers more and treat them better.
More from Bloomberg's opinion:
This column does not necessarily reflect the views of the editorial staff or Bloomberg LP and its owners.
Parmy Olson is a Bloomberg Opinion technology columnist. She is a former reporter for the Wall Street Journal and Forbes and author of “Supremacy: AI, ChatGPT and the Race That Will Change the World.”
More stories like this are available at bloomberg.com/opinion
Check out all the technology news and updates on DailyExpertNews. .
Must-have deals during Amazon Great Republic Day Sale 2025!
Get great deals on laptops, major appliances and kitchen appliances, gadgets, winter essentials and more. Celebrate Republic Day with Amazon's blockbuster sale and save big on top brands.
MoreFewer