There is a quiet colonial logic running through the latest wave of physical AI development. Human Archive, founded by Berkeley and Stanford researchers, is paying gig workers in India to wear camera-equipped caps and sensor devices, essentially renting their embodied experience to train robots that will, in all likelihood, replace other workers elsewhere. The throughline from surveillance capitalism to body capitalism is shorter than anyone wants to admit.
Physical AI and the New Labor Extraction Economy
The setup rhymes uncomfortably with a longer history of extracting raw material from the Global South to power Northern industry. Only now the raw material is movement, gesture, and spatial cognition. A 2026 paper on arXiv by Zhai et al., 'How Much Thinking is Enough?', found that reasoning-capable LLMs produce significant redundancy in their chain-of-thought outputs, raising questions about what kind of human cognitive labor is actually being absorbed versus discarded. Meanwhile, Imran and Bulathwela's work on AI tutoring blind spots shows that automated systems routinely misread the quality of human reasoning, seeing only the surface of a correct answer. These aren't separate problems. They are the same problem: machines learning the shape of human effort without understanding its texture. TurboFund's seed-stage AI investor list tracks exactly which funds are backing this physical AI wave, and the capital is moving fast.
Who Owns the Data When the Data Is You
Fast Company's reporting on wearable AI for professional athletes frames body-sensor tech as empowering and preventive. But the asymmetry is stark: NBA players have agents, unions, and contracts governing data use. Gig workers in Jaipur wearing sensor caps for Human Archive have terms of service. The body as training set is not a neutral technical decision. It is a political one, and the infrastructure being built now will determine who benefits from the embodied intelligence of billions of people who will never see a robot they helped teach.