Economists model the intelligence explosion question: 6x R&D growth bought only 2x capability gains
sudoraohacker · x · 2026-11-14
alexolegimas and lugaricano build a quantitative framework for assessing whether we're in an intelligence explosion (IE).
- METR reports frontier capability growth roughly doubled after October 2025, but annualized R&D spending grew 6x and technical staff 3x over the same window—indicating gains still come from exponentially rising inputs (scaling laws), not AI substituting for external inputs.
- The authors estimate several research production functions (building on Chad Jones, Aghion et al.) where a key parameter crossing a threshold signals an IE.
- The accompanying charts show a zero-breakpoint Jones model fits the METR trend (1.93x speedup), with estimated φ ≈ 0.76 and λ ≈ 0.948—below the φ=1 explosion threshold.
More from AGI Musings
- 80,000 Hours' Wiblin urges funding grassroots AI risk groups, sparking 'epistemics' drama — zetalyrae · 2026-11-14
- '2027 will be the year of principals': the agent vs. principal debate starts — chickefitz · 2026-11-14
- François Fleuret asks if AI is already being ring-fenced for a tiny elite — francoisfleuret · 2026-11-14
- "I made this completely without AI" flex is a dogmatic dead end, argues artist — Merzmensch · 2026-11-14
- The optimism-writing vibe shift in AI: leading with promise instead of peril lists — generativist · 2026-11-14
- After Outlook and FAA outages, warning that AI adoption lacks institutional guardrails — asusarla · 2026-11-14