I Work in Hollywood. Everyone Who Used to Make TV Is Now Secretly Training AI

The buried angle here is supply-side: AI companies are quietly solving their data quality problem by drawing on a pool of skilled, desperate workers who understand narrative, dialogue, and tone far better than generic crowdsource labor. That's not charity toward displaced creatives, it's a cost-efficient sourcing strategy that also happens to produce higher-signal training data.

This story doesn't connect directly to the Nvidia CUDA piece from May 11, which focused on infrastructure lock-in at the compute layer. But that's actually instructive: the CUDA analysis showed how AI's economic value concentrates at the infrastructure level, with software moats accruing to platform owners. This Hollywood story is the human-facing version of the same dynamic. The value created by creative labor flows upward to model developers and platform operators, while the workers supplying it remain in a gig arrangement with no accumulating stake. Together the two pieces sketch a consistent picture of where AI's structural advantages are consolidating.

Watch whether any of the major AI training platforms (Scale AI, Surge, or comparable contractors) begin publicly recruiting from entertainment industry guilds or alumni networks in the next six months. That would signal the informal pipeline described here is becoming an intentional sourcing strategy rather than an opportunistic one.