AI Futures Project: Intelligence Explosion in 2027, 5 Years of Progress in One Year
Hesamation · x · 2026-11-11
AI Futures Project (Eli Lifland, Daniel Kokotajlo, et al.) publishes a Q3.5 timelines update with far more aggressive conclusions than three months ago.
- The biggest driver is METR's recent report: Anthropic is now making AI progress 2x its historical rate (range 1.3–3x), with perhaps 80% of that acceleration attributed to AI itself
- Three months ago, even their most aggressive model gave <4% odds of 1.7x acceleration by now — making this revision itself surprising
- New forecasts: Superhuman AI Researcher (SAR) likely arrives in 2027 (absent deliberate slowdown); 35–65% chance Top-Expert-Dominating AI automating all white-collar work arrives before 2028
- If SAR is coming in 2027, they argue an industry-wide pause and international agreement become incredibly urgent, since no company is on track to align a SAR
Related event: AI 2027 Team Updates Timeline: Intelligence Explosion Now Expected in 2027(7 posts)→
More from AGI Musings
- BVSR: one equation to measure AI-scientist creativity, beating baselines across fields — james_y_zou · 2026-11-11
- Menlo Ventures: enterprise AI deployment spending hits ~$177B, 5.4x growth, agents dominate token use — ingliguori · 2026-11-11
- Microsoft 365 and Google Workspace have 4-5 years left as AI-native tools kill the document paradigm — signulll · 2026-11-11
- Design flipped: from brick-by-brick to chipping a sculpture out of marble in the AI era — round · 2026-11-11
- Security vendor POCs set to get far tougher in the agent-native era — _shreya_s · 2026-11-11
- Nine-country study: 51% now use conversational AI, but audiences don't want it making news — _FelixSimon_ · 2026-11-11