On its About page, the data-labeling company Scale AI states that it has paid more than $1 billion to its global community of contributors and has processed roughly 15 billion human decisions used to train AI models. The figures put a number on a usually invisible part of the AI supply chain: the human workforce that labels images, ranks model outputs, and annotates data so that “automated” systems can learn. Scale supplies this human-generated training data, reinforcement learning from human feedback, and model evaluations to leading AI labs, making the billions of small human judgments it coordinates a foundational input to modern AI.