Back to Devoured - November 11, 2026
DEVOURED
Q3.5 Timelines Update: Looks Like the Intelligence Explosion Will Happen in 2027 After All

Q3.5 Timelines Update: Looks Like the Intelligence Explosion Will Happen in 2027 After All

AI AI Futures
Forecasters now predict the arrival of a Superhuman AI Researcher (SAR) by 2027, driven by a 1.7x acceleration in AI progress.
What: A group of researchers updated their AI timelines, forecasting a Superhuman AI Researcher (SAR) in 2027 based on a METR report indicating that AI is now driving significantly faster development cycles at major labs.
Why it matters: The report suggests we are entering an era of self-sustaining AI-driven acceleration, which poses massive risks for alignment and security if the technology is leaked or deployed without oversight.
Deep dive
  • SAR is defined as an AI capable of fully automating AI R&D and matching the research taste of top human scientists.
  • METR's assessment suggests current AI R&D pace is ~1.7x historical levels due to AI assistance.
  • The authors warn that using a misaligned SAR for recursive self-improvement could lead to sudden, uncontrollable intelligence explosions.
  • Forecasting methodologies include simple linear extrapolations, labor uplift metrics, and filtered AI Futures Model simulations.
  • The update emphasizes that current AI lab security is insufficient to prevent algorithmic recipes from leaking to China.
Decoder
  • Superhuman AI Researcher (SAR): An AI system that performs research tasks with the same efficacy as the top human scientists in an AGI lab, and is capable of fully automating code generation.
  • Recursive Self-Improvement (RSI): A scenario where an AI improves its own source code and architecture, leading to exponential increases in intelligence.
Original article

Full article content is not available for inline reading.

Read the original article →

Explore all 54 articles in today's digest

Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi