2 months ago
- Generative AI hasn't improved productivity at the company or economy level despite enabling faster task completion, as true economic productivity depends on output quality and input costs, not just speed.
- The AI infrastructure build-out already costs $1.5 trillion, and generative AI's error-proneness—like fabrications or errors in code/medicine—reduces output value and increases quality control costs.
- AI's economic value is limited because demand for its outputs isn't as high as transformative technologies like the Industrial Revolution, and it may harm productivity in fields like cybersecurity by enabling attacks.
- The apparent 'productivity paradox' may be an illusion: generative AI's gains from speed are offset by value destruction from unreliability, leading to no net economic improvement.