Back to papers
other8.0 / 10

AI 2027: What Superintelligence Looks Like

Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean (AI Futures Project)

Abstract

Detailed scenario and forecasting report predicting AI agents will begin automating AI R&D by 2027, triggering an intelligence explosion and US-China arms race. Includes quantitative predictions, a takeoff model, and two branching endings (Race vs. Slowdown). One of the most widely discussed AI safety documents of 2025 on LessWrong and EA Forum.

Research area

ai alignmentalignmentpost-training
Published
2025
Source
other
View paper
Sign in to read and join the discussion.