Artificial intelligence is driving some of the biggest technological advancements of our time, from self-driving cars to facial recognition.
But as Axios reports, this technology is reliant on a growing, often low-paid sector of human workers, or “AI sharecroppers.”
Before an AI system can identify images on its own (a process called deep learning), it must be “trained” with millions of hand-labeled images.
This so-called “AI labeling” industry is projected to grow from $150m globally in 2018 to $1B by 2023 — and like many booming industries, it’s largely dependent on tedious, cheap labor.
Many of these jobs are outsourced to Southeast Asia and Kenya, where workers sit at computers labeling millions of images (stop signs, animals, vehicles) for as little as $2.50 per hour.
But as IEEE Spectrum reports, some of these jobs also find a home on US soil.
Alegion, based in Texas, promises to help veterans “find meaningful work in the new digital gig economy.” (Apparently, this means labeling aerial photos of cars and trucks for $7-15 per hour.)
Big-time benefits for a small-time cost
Though the work these labelers do is rote, repetitious, and often referred to as “unskilled,” the financial value AI companies reap from it is enormous.”The companies derive benefit over a long time, while workers are paid just once,” James Cham, a VC at Bloomberg Beta, told Axios. “They are paid like sharecroppers, making subsistence wages. The landowners get all the returns because of how the system is set up.”