AI Safety

djbinder's AI Economy Part 3: Doubling every 4-8 months by reoptimizing production recipes

Post-AGI economy could double every 4-8 months using only existing technology, not novel breakthroughs.

Deep Dive

This article extends the AI Industrial Explosion series by considering how production methods would change after AGI. Labor becomes nearly free, capital reproduces quickly, and interest rates rise sharply. The author applies four channels to US 2017 input-output tables: stripping out capital/services that accommodate humans, closing the gap between average and frontier plants, building cheaper short-lived capital, and shifting to more labor-intensive recipes. Additionally, by using the cheapest recipe from any historical era (1947 onward), the economy can grow even faster. Construction lags become more severe at higher growth rates, so the model uses sector-specific build durations and deployment lags for heavy industries.

The central estimate is a maximum self-reproduction rate between 1 yr⁻¹ and 2 yr⁻¹, corresponding to doubling times of four to eight months. This is roughly twice as fast as Part 1's one-year doubling. The savings-rate adjustment from Part 1 still applies, though resource depletion adjustments may differ with reoptimized recipes. The author notes that existing methods cannot push the rate much higher; Part 4 will explore improved technology. The piece draws on historical US input-output data and includes detailed appendices on construction lags and factor prices.

Key Points
  • Central estimate: economy doubles every 4-8 months (1-2 yr⁻¹ growth rate)
  • Twice as fast as Part 1's one-year doubling time, using only existing technology
  • Four reoptimization channels: strip human accommodations, close productivity gaps, cheaper short-lived capital, shift to labor-intensive recipes

Why It Matters

For professionals: this suggests AI-driven growth could far outpace current models, demanding rapid strategic adaptation.