I’ve been interested in AI risk for a while and my confidence in its seriousness has increased over time, but I’ve generally harbored some hesitation about believing some combination of short-ish AI timelines[1] and high risk levels[2]. In this post I’ll introspect on what comes
