FIR #508: Inside AI’s Human Raw Material Supply Chain

When workers lose their jobs, many turn to gig work to earn income while waiting for new opportunities. Increasingly, companies that hire gig workers are shifting from delivering food or sharing rides to creating content to train AI systems. This raises various communication and ethical issues. Neville and Shel explain what’s happening and discuss the implications in this short midweek episode. Links from this episode: The jobs AI can’t do – and the young adults doing themThousands of people are selling their identities to train AI – but at what cost?The gig workers who are training humanoid robots at homeGig economy becomes new AI training ground The next monthly, long-form episode of FIR will drop on Monday, April 27. We host a Communicators Zoom Chat most Thursdays at 1 p.m. ET. To obtain the credentials needed to participate, contact Shel or Neville directly, request them in our Facebook group, or email [e-mail]. Special thanks to Jay Moonah for the opening and closing music. You can find the stories from which Shel’s FIR content is selected at Shel’s Link Blog. You can catch up with both co-hosts on Neville’s blog and Shel’s blog. Disclaimer: The opinions expressed in this podcast are Shel’s and Neville’s and do not reflect the views of their employers and/or clients. Raw Transcript Shel Holtz Hi everybody and welcome to episode number 508 of For Immediate Release. I’m Shel Holtz. Neville Hobson And I’m Neville Hobson. Over the past few weeks, I’ve come across a set of stories that all point to something quite striking — not just how AI is evolving, but how it’s being built. Increasingly, the raw material behind AI isn’t just data scraped from the web. It’s us: our voices, our movements, our everyday lives, and increasingly, our identities. There’s a new layer of the gig economy emerging. We’ll explore this in just a minute. People are being paid, typically in small amounts, to record themselves walking down the street, having conversations, folding laundry, even just going about their day. That data is then used to train AI systems because those systems need examples of how people actually speak, move, and interact in the real world. In one case, delivery drivers in the US are being redirected to film tasks for robotics training. Platforms are turning existing gig workers like delivery drivers into distributed data collectors for AI. In another example, people are selling access to their phone conversations through apps that pay contributors to upload voice and text data. And in yet another, workers are strapping phones to their heads to record household chores so humanoid robots can learn how to move. The work is global, fragmented, and often invisible, with workers spanning Nigeria, India, South Africa, the US, and far beyond. Humans are no longer just users of AI — they are raw material suppliers. In China, there are even state-run centers where workers wear virtual reality headsets and exoskeletons to teach robots how to carry out everyday physical tasks. What we’re seeing is the rise of what you might call data labor, where identity itself becomes part of the work. There’s a clear driver behind it. AI companies are running out of high-quality training data. The open web isn’t enough anymore, and synthetic data has its limits. So the industry is turning to something else: real human lived experience. Because if you want a robot to understand how to load a dishwasher, navigate a room, or interact with objects, you need to see humans doing it at scale. But there’s an interesting contrast here. One of the stories highlights a 23-year-old in the US, a guy called Cale Mouser, who earns well into six figures repairing diesel engines. It’s something he’s developed great skill in doing. His work depends on judgment, experience, and problem solving in the real world — things that don’t easily translate into data. So while some people are being paid small amounts to generate data for AI systems, others like Cale Mouser are building highly valuable careers precisely because their skills can’t be reduced to it. And that contrast feels important. Because on one level, this new kind of work does create opportunity. For some people, especially in lower-income regions in the Global South, this is real income — paid in dollars, flexible and accessible. But there’s another side to it. Because what people are actually selling isn’t just time, it’s identity: their voice, their behavior, their presence in the world. And often once that data is handed over, it’s gone — permanently licensed, reused, repurposed, potentially in ways the individual never sees or understands. So you have this asymmetry: individuals earning small immediate payments while companies build long-term, highly valuable AI systems. Perhaps it’s a new version of the Mechanical Turk for the AI era. And that raises a deeper question. What does it mean when ...