Cheap Chinese AIs threatening frontier labs is a 'lump of labor fallacy,' argues Theo Jaffee
robleclerc · x · 2026-08-27
Theo Jaffee argues that the claim cheap Chinese AIs threaten frontier labs' business models — because people will use cheaper models for most tasks — is the lump of labor fallacy: there isn't a fixed pool of tasks. Better frontier models open up huge numbers of new tasks, some enormously valuable, like fully automated white-collar labor. In reality, frontier lab and hyperscaler revenues keep growing astronomically while frontier models stay months ahead of the best open-source options. Rob Leclerc adds a follow-up question: has anyone analyzed whether an AI trained by a lab is as effective on that lab's own codebase?
More from AGI Musings
- I Asked Grok Whether AI Capex Is a Bubble — Its Answer Was Sobering — BennyBop · 2026-08-27
- Temu 'Oai' listing sparks speculation of real OpenAI release — curious_vii · 2026-08-27
- Normies blindly trust AI outputs without quality checks — RachelVT42 · 2026-08-27
- AI Is Making Me Dangerously Intolerant of Unnecessary Process — moontripss · 2026-08-27
- Princeton's Narayanan at ICML: AI is augmentation, not automation — plenty of work left for humans — sayashk · 2026-08-27
- Large companies structurally unable to leverage AI for speed, small teams gain edge — antirez · 2026-08-27