Efficiency vs Capability in the Intelligence Explosion
Karthik Tadepalli · arXiv · 24 Oct 2026
Read original sourceOpen access
Where this sits in the debate
Direction of the finding as read from the abstract by an AI model (AI-written).
Key takeaway AI-written
Using a task-automation model of AI research, the paper shows that a self-sustaining acceleration of AI progress with fixed compute and labor requires both AI capability and AI efficiency to improve fast enough, and neither can substitute for the other. Calibrated to recent progress, capability looks fast enough while efficiency is on a knife-edge; in Monte Carlo simulations such an acceleration occurs in 49% of draws, and uncertainty about efficiency progress drives most of the uncertainty.
Abstract
AI progress can feed back into AI research in two ways. Capability improvements allow AIs to perform a larger share of AI research tasks, while efficiency improvements raise AI research output per unit of compute on tasks already automated. In a task automation model of AI research, I show that a software-driven intelligence explosion - a self-sustaining acceleration in AI progress with fixed compute and human labor - requires both margins to improve sufficiently fast, and neither margin can substitute for the other. To see whether these conditions hold, I calibrate both margins with data on recent AI progress. I estimate that capability progress is likely fast enough to support an SIE, while efficiency progress is on the knife-edge. In Monte Carlo simulations, an SIE occurs in 49% of draws, indicating that an SIE is plausible but uncertain based on historical trends in AI progress. The single largest source of uncertainty is the rate of efficiency progress, whose uncertainty is responsible for 45% of the uncertainty about whether an SIE occurs, more than all other parameters combined. Thus, better data on efficiency progress in frontier models is essential for understanding the likelihood of an SIE.
About this publication
Preprint. A preprint — posted publicly before or without journal peer review; findings may still be revised.
- Published
- 2026 (recently)
- Citations
- Not recorded for this source
Citation counts come from OpenAlex; none are recorded for this source — absence of a count is not absence of citations.
Related by theme, not necessarily by method or conclusion.
- Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality
SSRN · 2023 - Generative AI at Work
The Quarterly Journal of Economics · 2025 - The Impact of AI on Developer Productivity: Evidence from GitHub Copilot
arXiv · 2023 - The Productivity J-Curve: How Intangibles Complement General Purpose Technologies
American Economic Journal Macroeconomics · 2020
Overview assembled automatically from collected metadata. How this site is built.