P(Doom)=1%? Doom Debates — Michael Timothy Bennett on AI Risk

A sharp, formal clash over whether advanced AI makes human extinction likely. In this condensed 5‑minute version (original 2 hours), host Liron Shapira and guest Michael Timothy Bennett square off on AI risk, alignment, embodiment, and timelines. Bennett argues P(Doom) with AI is only about 1% by reframing intelligence as embodied adaptation and describing three optimization strategies (scale‑maxing, simp‑maxing, W‑maxing) that favor persistence over omnipotent optimization; Shapira presses instrumental convergence, fast capability gains, and bootstrapping routes to catastrophe. Listeners will come away with clearer keywords and concepts to explore—AI risk, alignment, orthogonality, embodiment, instrumental convergence, and governance—and a succinct map of the empirical and philosophical stakes. Listen now to get the key ideas in minutes.