When Nick Bostrom published Superintelligence in 2014, many of its ideas seemed closer to science fiction than an imminent technological challenge. Had someone recommended the book to me even a few years ago, I probably would have dismissed it as an interesting but highly speculative thought experiment. The rapid emergence of generative AI, followed by increasingly capable AI agents, completely changed my perspective. Curious to revisit the subject with fresh eyes, I finally picked up Bostrom's now-classic work. Despite being over a decade old, it remains remarkably relevant. Drawing on his background in philosophy, physics and computational neuroscience, Bostrom goes far beyond the technical aspects of AI to explore one of the deepest questions of our time: if humanity succeeds in creating a superintelligence, how can we ensure that it remains aligned with our interests? These are the ideas that stayed with me after six weeks of reading.
