AI 2027 team's superintelligence scenario tops a list of in-depth AI futures

RosieCampbell · x · 2026-09-12

A user rounds up the 5-10 in-depth AI future scenarios written over the years, pointing to AI 2027 as the easiest to reference. Written by Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland and Romeo Dean, it predicts superhuman AI's impact over the next decade will exceed the Industrial Revolution, with both a "slowdown" and a "race" ending. The authors stress the goal is predictive accuracy rather than advocacy, note 2027 was only their modal estimate, and invite debate via prize-funded alternative scenarios. Also cited: scenario essays by Paul Christiano and Gwern Branwen.

Original post →

More from AGI Musings

AGI Musings channel →