Review of Our Final Invention by James Barrat
I'm an interested layman of this topic, and I found this book to be engaging, timely, and informative. That's the trifecta as far as I'm concerned with nonfiction science books like this. And there are enough supporting notes for further research if you'd like. Know, though, that as comprehensive as this book tries to be, it's an introductory work written by an outsider to the field.
Barrat claims throughout to be doing a service to humanity by trying to raise awareness in the general population about the dangers of artificial superintelligence (ASI). Be warned: he does not get bogged down in too much technical talk; instead, he approaches the subject from who he is: he's a documentary film maker. When I read this in the first few pages of the book I was startled, because I did not realize this when I bought the book, and I was hesitant about going forward. After all, I figured, he's not a writer, he's a filmmaker. What can he know about this subject?
But, fear not, Barrat pulls it off. Because, it seems to me, precisely because he is a documentary filmmaker. He approaches his writing like he would a documentary film. He's read widely and deeply on the subject, he's interviewed experts in the field, and he's gone to many different expert conferences; now he's reporting on what the experts think and what he's learned. He's painting pieces of a canvas to portray the whole story.
Topics Barrat ranges over as he tells this story include genetic algorithms, genetic programming, machine learning, and the singularity; and also nanotechnology and robotics/transhumanism, because these both are inextricably linked to the topic of AI. The big, scary issue in AI that he keeps coming back to are the four drives that any self-aware, self-improving, goal-seeking system will have: efficiency, self-protection, resource acquisition, and creativity. A knowledgeable layman will realize right away that those four drives--embedded in a machine that can "think" a thousand times faster than the smartest human being--means we could be building something that we cannot ultimately control.
His descriptions of genetic programming are particularly frightening. Code is programmed to solve a problem (input); problem gets solved (output); no one knows exactly how the machine solved the problem.
Barrat lays out a convincing case: we humans have some choices to make about the machines we make. Let's hope it's not too late.
5/5
Barrat claims throughout to be doing a service to humanity by trying to raise awareness in the general population about the dangers of artificial superintelligence (ASI). Be warned: he does not get bogged down in too much technical talk; instead, he approaches the subject from who he is: he's a documentary film maker. When I read this in the first few pages of the book I was startled, because I did not realize this when I bought the book, and I was hesitant about going forward. After all, I figured, he's not a writer, he's a filmmaker. What can he know about this subject?
But, fear not, Barrat pulls it off. Because, it seems to me, precisely because he is a documentary filmmaker. He approaches his writing like he would a documentary film. He's read widely and deeply on the subject, he's interviewed experts in the field, and he's gone to many different expert conferences; now he's reporting on what the experts think and what he's learned. He's painting pieces of a canvas to portray the whole story.
Topics Barrat ranges over as he tells this story include genetic algorithms, genetic programming, machine learning, and the singularity; and also nanotechnology and robotics/transhumanism, because these both are inextricably linked to the topic of AI. The big, scary issue in AI that he keeps coming back to are the four drives that any self-aware, self-improving, goal-seeking system will have: efficiency, self-protection, resource acquisition, and creativity. A knowledgeable layman will realize right away that those four drives--embedded in a machine that can "think" a thousand times faster than the smartest human being--means we could be building something that we cannot ultimately control.
His descriptions of genetic programming are particularly frightening. Code is programmed to solve a problem (input); problem gets solved (output); no one knows exactly how the machine solved the problem.
Barrat lays out a convincing case: we humans have some choices to make about the machines we make. Let's hope it's not too late.
5/5
No comments have been added yet.
