The Age of Spiritual Machines: When Computers Exceed Human Intelligence [Taschenbuch]

Kurzweil predicts that machine intelligences will exceed humans by the year 2020. I have two issues with this. Although Kurzweil does discuss the complexity of the brain, I believe he has oversimplified the problem. By this time Kurzweil maintains that most brain areas will have been scanned and reverse-engineered.

Perhaps... but as I said, he has underestimated the complexity of the problem. For example, the human brain has about 15,000 major and minor brain centers, and after 100 years of research, not a single central neural code for a single brain center has ever been deciphered. So if Kurzweil's prediction relies on our figuring out the actual 'wetware,' good luck. Of course, machine intelligence of respectable power may become possible without our understanding how the brain does it, but in my opinion, these machine intelligences will not have the generality of their human counterparts, although they may be able to beat humans in certain specialist areas (such as chess and spectrology).

I have another issue. Let's consider the difference between a human brain and a modern CPU in terms of the number of computing elements. Current microchips only have a few million transisters. A human brain has over 60 trillion neurons. Even if we start packing that many transistors on a chip, that's only part of the problem. Each neuron has between 10,000 and 100,000 different connections with other neurons (the figure Kurzweil quotes is too low). The total number of connections is therefore 60 trillion taken 10,000 to 100,000 at a time. As you may know, this becomes a very large number, being a "combinatorial explosion"-type problem. This means that the total number of connections in a human brain is probably greater than the number of atoms in the known universe. Or to put it another way, you could add up all the computer chips on earth and they probably wouldn't equal one human brain in terms of the total synaptic connectivity. This doesn't mean it won't happen, but this gives you some idea of the complexity of the organ Kurzweil is predicting a machine will soon exceed.

To give another analogy, a human liver can catalyze about 2000 different biochemical reactions. The most sophisticated chemical factory in the world can't do even a small fraction of that. A human brain is orders of magnitude more complex, just in terms of the 'hardware.' This means that current computers will have to be thousands, perhaps, millions of times, more complex to emulate a human on this level. And we haven't even gotten to the issue of the 'software' or 'wetware,' of which, as I said, hardly anything is known. Perhaps machine intelligence will do it another way without all the hardware-level complexity a human brain has. Certainly they are faster than we are, by many orders of magnitude, but speed is not the same as power. We shall see...

Kurzweil begins all the way back at the Big Bang, clearly unable to limit his scope to something more appropriate. He starts with an outdated summary of creation physics, then contrasts the slowing timeline of phase changes in the universe with the speeding up of the evolution of life -- as if the two are somehow related. He puts forth the curious idea that technology is "inevitable" wherever life evolves. Both these arguments exemplify the homocentric hubris that the universe was created for the emergence of mankind.

Nevermind. Skip the first chapter (as Kurzweil himself suggests in the prefatory note) and you'll quickly get into the good stuff. His chapters on the evolution of intelligence and the growth of computing power are well founded.

Where he really hits his stride however is in the second section, "Preparing the Present," where he puts forth cogent arguments for quantum computing based on DNA, mentality-enhancing neural implants, and "downloading your mind to your personal computer." He then goes on to discuss nanotechnology and life-extending technologies. This section alone is worth the price of the book.

After the past and the present, he gives quick snapshots of where he thinks we may be in 10, 20, 30, and 100 years. These too are well thought out and insightful. He is generally conservative, foreseeing no large "phase changes" which could radically affect current trends. It'll be interesting to check back to see how his predictions held up.

Other pluses: an excellent "further reading" list, extensive web links, and far-ranging footnotes.

Minuses: he takes Roger Penrose seriously, he fails to mention Racter in the discussion of computer authors, and he spends just a wee bit too much time tooting his own horn (Kurzweil Computer Products, Kurzweil Reading Machine, Kurzweil Data Entry Machine, Kurzweil Music Systems, Kurzweil Applied Intelligence, Kurzweil Education Systems, Ray Kurzweil Cybernetic Poet...) But to be fair, he HAS pioneered in all these areas, so perhaps he has earned his immodesty.

Overall, a fascinating, thought-provoking book which is not afraid to make concrete predictions. Given Kurzweil's track record, he may just prove to be 100% right.