Summary
Rest of World review of The AI Con (book). Documents how the ImageNet project (2009) — a massive image labeling effort by crowdworkers — was the foundational event that made deep learning possible, establishing crowdwork platforms as the invisible infrastructure of AI. The book details worker exploitation in inflated Venezuela and refugee populations, subcontracting liability evasion, arbitrary account suspensions, and profit concentration. Argues ghost work could become sustainable employment with stronger protections.
Key Points
- ImageNet (2009): Stanford’s massive labeled dataset was built by crowdworkers; it launched modern deep learning — making human data labelers the foundational infrastructure of the AI era
- Workers recruited in economically distressed geographies: inflated Venezuela, refugee camps
- Subcontracting structures: tech companies → agencies → workers; each layer provides plausible deniability
- Algorithmic management: account suspensions without explanation; no appeal process
- Profit flows upward through the chain while costs (financial and psychological) concentrate at worker level
- Authors’ argument: ghost work could be good employment with better protections; the exploitation is a choice, not a structural necessity
Newsletter Angles
- ImageNet as the original sin: If the deep learning revolution was built on labeled data from crowdworkers, then every downstream AI system — every chatbot, image classifier, recommendation engine — has ghost labor at its foundation. This isn’t peripheral to AI; it’s constitutional.
- The “could be good” argument: Hanna, Bender, and Chandran make the case that ghost work is structurally necessary but exploitation is optional. This reframes the narrative from “automation will replace workers” to “the exploitation is a deliberate design choice that could be redesigned.”
- Venezuela and refugee populations: The geographic targeting of AI data work at populations in extremis — hyperinflation, displacement — is exploitation of crisis conditions for corporate profit extraction. This is a distinct dimension not captured by the general “Global South” framing.
Entities Mentioned
- Amazon — AMT as foundational platform
- Mechanical Turk Pattern — the book’s central argument; ImageNet as origin point
Concepts Mentioned
- Mechanical Turk Pattern — ImageNet as the founding event; subcontracting as evasion structure
- Leverage Erasure Through Automation — zero leverage for workers in algorithmic management systems
Notes
Review of The AI Con by Alex Hanna, Emily M. Bender, and Rina Chandran. The book is the primary source; this is a secondary review. The ImageNet founding narrative is the most distinctive contribution — connecting 2009 crowdwork to the current AI moment.