This is Bostrom’s strongest chapter, measured by predictive accuracy over 12 years. The racing dynamics he warned about in 2014 materialized spectacularly at OpenAI in 2024: the Superalignment team was dissolved, co-founder Sutskever left, alignment lead Leike departed saying safety had ‘taken a backseat to shiny products,’ and the company fully transitioned from non-profit to for-profit. The international governance response has been real but weaker than Bostrom proposed: the Bletchley Declaration (28 nations, November 2023) and EU AI Act (August 2024) exist but lack enforcement teeth. The one clear miss: the AI market is multipolar (OpenAI, Google, Anthropic, Meta, xAI, Chinese labs), not winner-take-all. Bostrom was right about the dynamics, wrong about the market structure.
So what
Chapter 14 is the most dramatically vindicated chapter of the book. The AI racing dynamics Bostrom predicted have played out almost exactly at OpenAI, with safety teams dissolved and leadership departing over commercial pressures. International governance has emerged faster than expected (Bletchley Declaration, EU AI Act, Biden EO) but remains weaker than Bostrom advocated. The one major miss is the winner-take-all prediction: the AI market is strongly multipolar, with no single decisive actor.
Verdict
MostlyAccurate