FRI report turns the recursive-self-improvement debate into forecastable questions

SchoeneggerPhil · x · 2026-11-13

Philip Tetlock shared and commented on a new ResearchFRI report: as uncertainty grows over how close we are to recursive self-improvement (RSI), two things likely rise too — noisier forecasts (volatility) and more polarized opinions (name-calling). Under information and emotional overload, he argues, signal detection degrades and debates get dumber exactly when we need them smarter.

The FRI report notes that AI companies increasingly say they may soon reach RSI, which could sharply accelerate capability progress inside labs before outsiders notice. It reviews the evidence, converts key disagreements into forecasting questions, and gives initial LLM forecasts.

Original post →

FRI report turns the recursive-self-improvement debate into forecastable questions

More from AGI Musings