AI’s Dark Output: The hardest measurement problem in history
- Source
- SemiAnalysis
- Time
- 11:47 PM
- Weight
- 94/100
Economists and data scientists are increasingly concerned that traditional macroeconomic indicators, such as Gross Domestic Product (GDP), are failing to capture the true value generated by artificial intelligence. Termed "Dark Output," this phenomenon describes AI-enabled economic value that is either invisible or distorted in national accounts.
Unlike the Industrial Revolution, where tangible goods like steel and screws were easily counted, AI's primary contributions occur within the service sector—an area where productivity measurement has historically been difficult due to the lack of standardized units of "mind power" or service output. Current statistical frameworks primarily track visible costs, such as GPU procurement, electricity consumption, and token expenditures, while failing to record the value of the tasks performed.