Top critical review
3 people found this helpful
on 22 July 2012
It saddens me to see so many glowing reviews of this book.
My background: 26 year old Computer Science BA/PhD from the University of Oxford, UK. Many-years-long fascination with neural networks, genetic programming etc, and implementation of some of the most modern complex techniques in aforementioned fields.
After reading the recommendations and reviews of this book, I was expecting something revelatory, inspirational and scientific. This book is, sadly, none of these things.
Steve Grand's book contains 15 chapters. 13 of these are self-indulgent narcissistic ramblings with superbly information deficient prose such as:
"We have uncovered the most important law of nature, and it is this: Things that persist, persist. Things that don't, don't."
These chapters combined feel like a botched attempt at Godel Escher Bach stardom, not only falling far short but failing to leave the ground at all.
The remaining 2 chapters are the closest to describing the Creatures implementation that you'll get. However these too are shockingly waffle-rich and information-poor. There's a reason for this: the "chemistry" Steve implements is crude and irrelevant to the creatures' successful mental evolution, and the "unique brain" he describes is nothing more than a simple neural network with a finite number of neurons such that those firing the least over time get disconnected and reconnected in a new, random position. The system he describes is essentially the world's worst implementation of genetic algorithms and genetic programming.
At time of writing, Steve has a video on kickstarter, a "make my project a reality by giving me money" site where you can advertise projects you need money to complete. He's raised money for "Grandrions", what he's publicly announced as a groundbreaking AI project. After researching this project, I can tell you several facts: (i) The kickstarter video is even more embarrassing than this book (ii) Grandrions is equally unremarkable, from the first video reveal it appears to be a software simulation of the "learning like a human" robots we see ubiquitously in modern science news, nothing unique or novel in any way, and (iii) he's shockingly behind schedule, despite accruing over twice the amount of funds he'd set as a target on kickstarter. The fact Steve has been awarded an OBE is an indictment of the intellect at the core of our government and the Queen's administration.
Go read about Gene Expression Programming by Candida Ferreira, An Introduction to Genetic Algorithms by Melanie Mitchell, or almost any Neural Networks text out there to learn about "real" AI and efficient machine learning. Don't waste your time on this dross.