Reviews & Press
Book explains limits of AI, wonders of human brain
By Lynn Yarris
Special to the Mercury News
Sun, Oct. 17, 2004
One of the hits at the movies this past summer was "I, Robot," starring Will Smith, in which a rogue band of humanoid robots threatens the human race. "I, Robot" was set in the year 2035. Is it possible that artificial intelligence (AI) will be that advanced in the next 30 years? Not if we continue down our current path of development, according to the man who was the creative genius behind the PalmPilot and the Treo smartphone.
In "On Intelligence," Jeff Hawkins takes a detailed look at how the human brain works, compares this to how AI currently works and concludes that our machines will never get there from here.
"Many people today believe that AI is alive and well and just waiting for enough computing power to deliver on its many promises," Hawkins says. "I disagree. AI suffers from a fundamental flaw in that it fails to adequately address what intelligence is or what it means to understand something."
Silicon Valley legend
Hawkins is a computer engineer, not a neuroscientist, but he's not just any computer engineer. He's a Silicon Valley legend, widely viewed as the father of handheld computing. Understanding the human brain and designing truly intelligent machines has been a lifelong passion for him, and he serves as director of the Redwood Neuroscience Institute, a non-profit research organization aimed at creating mathematical models of memory and cognition. His co-author, Sandra Blakeslee, is a veteran science writer for the New York Times who co-wrote a previous book about the brain, "Phantoms in the Brain: Probing the Mysteries of the Human Mind."
Together, Hawkins and Blakeslee make a terrific team that has produced a stellar book. If you have any curiosity at all as to how your brain works, read the "On Intelligence" prologue and I promise you'll be hooked. Hawkins' basic premise is that the human brain is a "memory-prediction system" for modeling the world around it. The brain stores and recalls experiences, relates them to other experiences and uses the memories of these accumulated experiences to make predictions. This ability to predict is the heart and soul of human intelligence, he says. It can also help explain phenomena such as consciousness, imagination and prejudice, or to answer such elusive questions as, What is reality?
"Why can a six-year-old hop gracefully from rock to rock in a streambed while the most advanced robots of our time are lumbering zombies?" Hawkins asks. "Why are three-year-olds well on their way to mastering language while computers can't despite half a century of programmers' best efforts? Why can you tell a cat from a dog in a fraction of a second while a supercomputer cannot make the distinction at all?"
The answer, as Hawkins and Blakeslee demonstrate, is that the human brain doesn't compute answers to problems; it retrieves answers from memory. While it takes a great many steps to compute something, it takes only a few steps to retrieve it from memory. The seat of human intelligence, where all this memory storage and retrieval takes place, is the neocortex, the pinkish- gray mass that constitutes about 85 percent of the human brain's total weight and looks like a cauliflower, with its numerous ridges and valleys.
Hawkins and Blakeslee do an outstanding job of explaining the organizational architecture of the neocortex and the intricate way in which its six layers of cells are connected to one another, as well as to other cells in the brain. Through a series of vivid, easy-to-understand examples, they show why our brains may be data-processing tortoises compared with computer hares, but human intelligence can easily cross finish lines that today's AI can't even find.
"The neocortex isn't made of superfast components, and the rules under which it operates are not that complex," Hawkins says. "However, it does have a hierarchical structure that contains billions of neurons and trillions of synapses. If we find it hard to imagine how such a logically simple but numerically vast memory system can create our consciousness, our languages, our cultures, our art, this book, and our science and technology, I suggest it is because our intuitive sense of the capacity of the cortex and the power of its hierarchical structure is inadequate."
As to the question of whether we can build truly intelligent machines, Hawkins believes the answer is yes, but those machines won't be the humanoid robots we're used to seeing in films like "I, Robot." For one thing, he thinks it would be impractical: A robot butler would be expensive and not very helpful. For another, without the emotional element, through which all our experiences are filtered, a robot would never be even remotely humanlike. Hawkins offers his own intriguing ideas as to what truly intelligent machines might be like, but to describe them here would be like giving away the ending to a novel.
A little more than three years ago, my wife suffered a stroke from which she is still recovering. Since then, I've read dozens of books about the human brain and how it works. "On Intelligence," with its chatty style and cross-references to the computer technology so ubiquitous in our lives today, is far and away the best.