Realizing these predictions requires that science deliver a computer that can think. That’s the premise that Kurzweil sets out to prove in his latest effort, “How to Create a Mind.” He argues that the brain’s fundamental building block for intelligence has been discovered by neuroscientists, that the algorithm for intelligence has been observed in nature and independently invented by artificial intelligence researchers, and that the steady progress of Moore’s Law will produce a computer fast enough to simulate an entire human brain by 2020. That wish is ultimately an appeal for a continuation of technological progress — humanity should create an intelligent machine unless something unforeseen stops us from doing so.
Kurzweil is at his best when he presents the reader with his “thought experiments on thinking.” For example, he asks you, the reader, to recite the alphabet. Next he suggests that you recite the alphabet backward. Most people can easily do the first but have a hard time with the second. This proves, he writes, that memories are stored as sequences of patterns that can be accessed only in the order in which they are remembered. Kurzweil presents similar experiments that he says establish that knowledge is stored in the brain as a series of hierarchical patterns, and that much of what we call “thinking” is really just pattern-matching and pattern-synthesizing. Of course, these simple thought experiments don’t really prove anything, but they are entertaining.
The next two chapters present Kurzweil’s misnamed “Pattern Recognition Theory of Mind (PRTM)” and delve into the anatomy of the human brain. PRTM is not a theory because it can’t be tested. For example, Kurzweil argues that neuroplacticity, the ability of one part of the brain to take on the functions of another that’s damaged, implies that different parts of the brain must use “essentially the same algorithm” to perform their computations. He then cites some recent neurological research to argue that this algorithm must run on some kind of neural “module,” which he says consists of about 100 neurons, and that there are roughly 300 million of these modules in each of our brains. That’s too big a conceptual jump for many of Kurzweil’s detractors, who say that the brain is likely to have many more secrets and algorithms than the ones Kurzweil describes. Over the next three decades we’ll see who is right.
Later chapters discuss scientists who are working to simulate a brain, briefly retell the history of computer science and present critiques of artificial intelligence from some of the field’s greatest detractors. It’s an eclectic collection, perhaps better suited to a dinner party or a TED talk than a scholarly effort; it’s also a bit disorganized. The arguments about the nature of consciousness are interesting, although Kurzweil has presented many of them before. His recipe for creating a mind, then, is to build something that can learn and then give it stuff to learn. That, after all, is what parents do when they conceive and raise children. But this is not “the secret of human thought” that Kurzweil promises in the book’s subtitle.
Sadly, Kurzweil’s in-book autobiography, repeated mention of his company’s products and snipes at his detractors come off as blatant self-promotion. This book would have benefited from a strong edit — perhaps in a few years there will be a program that Kurzweil trusts to critique his work. As it stands, much of the warmth and humanitarianism that are so evident in his talks are lost in this written volume.
writes and researches information technology. He is the author of 14 books, including “Architects of the Information Society: Thirty-Five Years of the Laboratory for Computer Science at MIT.”