State media framing here matters: Beijing is positioning data annotation as a durable job category even as frontier labs push toward synthetic and self-supervised pipelines that need less raw human labeling. If China's annotation shops are scaling headcount rather than shrinking it, that's a data point against the thesis that annotation demand collapses once base models mature—though it also raises the question of whether Chinese labs like those competing with Baidu, Alibaba, and domestic AI shops are simply buying more human-labeled tokens at lower cost per unit.
Watch whether this signals a wage floor forming in China's annotation labor market even as US buyers shift spend toward RLHF and expert-annotation tiers.
China's data annotation industry highlights how employment evolves with tech advancement
