Superintelligence: Paths, Dangers, Strategies

The breakthrough, introductory book on artificial general intelligence that begs to be read slowly and in real depth. This introduces you to so many of the potential ills and triumphs that could result from AGI during the near, medium and long-term future, but the main message I took from it is that, due to the inherent nature of an agent that by definition exhibits intelligence above that of our own, understanding the real tree of potential outcomes we face is really not a task cognitively possible with the framework of our (current) human intelligence. As Bostrom writes in his updated postscript for this version of the book, this is the critical point that most forget to factor in, whether in criticism of his argument or in that of anyone else who dares write about this subject. I highly recommend the book; while not the quickest to get through, it imparts to the reader so many novel, useful, and essential concepts.