Superintelligence by Nick Bostrom reads a bit like a textbook. And unless you’re already passionate about the topic, it requires patience and perseverance to get through it. While the content is complex, the structure is simple and straightforward, making Superintelligence easy to follow. I’m not an AI expert; I picked up this book to learn. And that’s exactly what I did.
Before I started reading Superintelligence, I knew enough about AI to understand its transformative potential and dangers under the control of humans—not because we’re inherently evil (we’re not) but because we’re human: we often fail to educate ourselves on technologies or monumental shifts in our world or we fail to fully think through the implications and potential outcomes of them.
Bostrom opens the book with this sober declaration:
“In this book, I try to understand the challenge presented by the prospect of superintelligence (i.e., machine brains that surpass human brains in general intelligence), and how we might best respond. This is quite possibly the most important and most daunting challenge humanity has ever faced. And—whether we succeed or fail—is probably the last challenge we will ever face.”
Fear does not characterize the book, however; a good dose of straight talk, research and data do.
My favorite chapter outlines the potential paths to superintelligence:
1) Artificial Intelligence: a program developed to simulate the human mind
2) Whole Brain Emulation: modeling or copying the structure of a real brain
3) Biological Cognition: enhancing the function of the human brain through modeling and AI techniques
4) Brain-Computer Interfaces: implants to the brain
5) Networks and organizations: linking human minds via artifacts and bots
This chapter alone is worth the price of the book because it demystifies—at the most basic level—the “black box” out there that we call AI. I’ll never read an article on artificial intelligence through the same lens again. He makes the point (effectively) that with so many potential paths to superintelligence, we really shouldn’t pooh- pooh its reality. Or its potential hazards if misused.
As the title implies, Bostrom goes on to address AI dangers and strategies, including technology strategy, enablers, and collaborative approaches.
In full disclosure, my mind strayed often from the heft of the content. But that’s okay. As an “audit student”, I don’t have to go from 0 to 60 in 1.7. And neither do you. Read it once if you want to get a “lay of the AI land.” Read it twice if you want to understand a little more about the soil. This is a great overview written by an expert in the field.