Debate Rekindled Over "Slow Takeoff": Yudkowsky vs. Christiano Revisited

The AI safety community has erupted in a fresh round of debate over Paul Christiano's definition of "slow takeoff," sparked by recent math-capability events that some view as a test of the related predictions.

Confirmed

• Christiano had operationalized "slow takeoff" as a path where "world GDP doubles in four years, then again in one year..." — a prophecy that Yudkowsky (@allTheYud) notes has yet to materialize.
• Katja Grace and Yudkowsky clashed directly: Grace pressed on what exactly about the "slow takeoff" claim has been falsified; Yudkowsky responded that if proponents intended to include today's upheaval within "slow takeoff," they had an obligation to state that explicitly and concretely in advance.
• jessicata used the moment to recommend rereading the classic 2021 LessWrong debate between Yudkowsky and Christiano on takeoff speeds as a framework for evaluating current AI predictions; a cited AI analysis concluded "Yudkowsky was right on the math, Christiano right on the economics."
• jan Kulveit pushed back against anyone "celebrating victory" over this definitional dispute and criticized the "slow vs. fast takeoff" terminology itself as having long been muddled; he believes current reality is closer to Christiano's continuous-takeoff view, but that AI sharpness and recursive self-improvement (RSI) risks are greater.

Why it matters

• At its core, this dispute is about how to judge whether AI predictions succeed or fail: vague definitions make "falsification" impossible, and Yudkowsky stresses that claimants bear responsibility for offering clear, testable statements.
• Kulveit's position offers a middle-ground perspective: the continuous-takeoff framework broadly holds, but sharpness and RSI risks exceed its traditional assumptions — with direct implications for risk assessment.
• The resurfacing of the 2021 debate shows the community is systematically revisiting existing prediction frameworks to calibrate judgments about current capability progress.

2026-10-07 ~ 2026-10-08 · 5 related posts

• [source] Yudkowsky: if 'slow takeoff' included today's math breakthroughs, proponents should have said so loudly — allTheYud · 2026-10-07
• [source] Katja Grace presses Yudkowsky on what exactly was falsified about slow takeoff — KatjaGrace · 2026-10-07
• [source] Revisiting the 2021 takeoff debate: AI analysis says Yudkowsky got the math, Christiano the economics — jessi_cata · 2026-10-08
• Yudkowsky reflects on Christiano's 'slow takeoff': the four-year GDP doubling anchor that failed — repligate · 2026-10-08
• jan Kulveit: reality tracks Paul Christiano's continuous takeoff so far, but with more spikiness and RSI risk — teortaxesTex · 2026-10-08