Doom Debates: Alex Turner on P(Doom), DeepMind Exit, and AI Misuse Risk

A former DeepMind AI safety researcher says the biggest danger may not be technical misalignment, but how governments and companies use powerful AI. In this concise summary of Doom Debates, Liron Shapira debates Alex Turner, who explains why he resigned from Google DeepMind, how his thinking evolved from LessWrong and Yudkowsky-style doom scenarios, and why he now estimates P(doom) at 25–30%. This shorter recap distills the original full-length discussion into a quick listen, highlighting their arguments about AI alignment, instrumental convergence, corrigibility, autonomous weapons, and whether market and social forces can keep advanced systems in check. You’ll also hear Turner’s view on GPT-2-era fears, the “Icarus curve,” and why he thinks assurance contracts could help coordinate safer AI policy. Listen now to get the key ideas in minutes.