The Human Supply Chain of AI

Algorithms do not learn. They are taught.

Artificial Intelligence is not magic. It is manufacturing.

We talk about AI as if it were a naturally occurring resource—mined from data lakes and refined by compute power. But this metaphor erases the most critical component of the assembly line: the human.

“Behind every ‘autonomous’ system is a human worker labeling the reality that the machine will eventually mimic.”

My research at UCLA Anderson focuses on this invisible infrastructure. I treat the training of AI models not as a computer science problem, but as a Supply Chain problem. We have raw materials (unlabeled data), bottlenecks (human attention), inventory costs (latency), and quality control issues (bias).

Just as the industrial revolution required us to study the factory floor, the AI revolution requires us to study the Crowdwork Platform. This is where the work happens. It is where the ethics of fair pay collide with the mathematics of estimation.

We cannot optimize what we refuse to see.