Will AI modestly increase productivity, restore rapid twentieth-century-style growth, or produce historically unprecedented economic expansion?
Acemoglu's task-based estimate is ~0.7% total TFP over a decade; Epoch-style growth models argue full automation implies growth rates without historical precedent. The disagreement is about substitution depth, not arithmetic.
View on the map → · Open in Browse →
The disagreement got quantified: a large forecasting survey puts median growth expectations near historical rates even as new models specify exactly when automated R&D turns explosive. The sharpest new claim is measurement: quality-adjusted AI output may be growing over 2,000% a year while remaining nearly invisible in GDP.
Structured survey of economists, forecasters, and AI-company staff: median forecasts near 2.5% baseline growth, ~4% under rapid AI — disagreement driven by beliefs about effects, not timelines.
Growth becomes explosive if the combined strength of technological and economic feedback loops overcomes diminishing returns.
Calibrated conditions under which automated AI R&D yields superexponential growth despite production bottlenecks.
we are not 100 million times more productive
Argues weak links — tasks machines cannot perform — have historically capped gains from massive increases in computing power.
Nominal AI GDP near $250B in 2025, but quality-adjusted output growing over 2,000% a year — invisible in GDP because falling prices offset quality gains.
2025's strong GDP with weak job growth is not yet evidence of an AI productivity boom: noisy quarterly data, revisions, and composition effects.
Maps the spread between 0.7% and 18% growth forecasts onto three testable disagreements: automation limits, institutional friction, automated innovation.
A markets-side skeptical case: sustained disruption requires several simultaneous extreme conditions; AI may mainly offset aging and deglobalization headwinds.