Hundreds of Los Angeles residents are earning money by strapping head- and hand-mounted cameras to their bodies while performing ordinary household chores — washing dishes, scrubbing toilets, making coffee — generating the embodied motion data that humanoid robotics companies cannot get anywhere else. As reported by Nilesh Christopher of the Los Angeles Times, platforms like Instawork are acting as labor intermediaries, deploying on-the-ground managers to distribute equipment and assignments at casual locations like downtown coffee shops. Workers such as Salvador Arciga, a veteran of the L.A. gig economy, earn roughly $80 for two hours of footage. Unlike the text and image data that trained large language models, physical AI systems need intimate, first-person recordings of natural human movement that cannot be easily synthesized — and gig workers are currently the only reliable source.

Goldman Sachs forecasts the humanoid robot market could reach $38 billion by 2035. Grand View Research projects the broader data collection and labeling market will hit $17 billion by 2030. Robotics firms including Figure AI, Tesla (via its Optimus program), and Dyna Robotics are the primary consumers of this training data, and a new tier of data infrastructure companies — Scale AI, Encord, Micro1, and Sunain — has emerged to manage collection and annotation pipelines at scale. Encord raised $60 million in February 2026 after its physical AI operations revenue grew tenfold in a year; Scale AI has amassed 100,000 hours of robotics footage; and Micro1 employs 1,000 workers across 60 countries recording household tasks using custom wrist cameras.

Workers collecting this data are, in effect, training the systems designed to automate the physical and service labor they depend on. <a href="/news/2026-03-14-ai-companion-chatbots-hidden-labor">The displacement is already visible</a>: Serve Robotics' AI-powered delivery robots have pushed out human drivers while simultaneously creating lower-value data-collection roles — the same substitution pattern, compressed into a single company. Rachel Kwon, a labor economist at UCLA who studies gig work in the AI supply chain, says the dynamic will only accelerate. "You're paying people to build the automation that eliminates their job category," Kwon said. "The only variable is the timeline." Workers have little legal recourse. Body motion capture data sits in uncertain territory under Illinois' Biometric Information Privacy Act, though Washington State's My Health My Data Act explicitly covers gait patterns and behavioral characteristics, and California's CPRA extended full data rights to independent contractors as of January 2023. No class action specifically targeting body motion capture data collected for robotics training has been filed as of early 2026 — leaving workers selling their movements at $80 a session with almost no established legal ground to stand on.