For starters, you should realize that as soon as a computer achieves a level of intelligence comparable to human intelligence, it will necessarily soar past it. A key advantage of nonbiological intelligence is that machines can easily share their knowledge. If I learn French, I can't readily download that learning to you. My knowledge, skills and memories are embedded in a vast pattern of neurotransmitter concentrations and interneuronal connections and cannot be quickly accessed or transmitted.
But when we construct the nonbiological equivalents of human neuron clusters, we will almost certainly include built-in, quick-downloading ports. When one computer learns a skill or gains an insight, it will be able to share that wisdom immediately with billions of other machines.
Let's consider the requirements for a computer to exhibit human-level intelligence, by which I include all the diverse and subtle ways in which humans are intelligent--including musical and artistic aptitude, creativity, the ability to physically move through the world and even to respond to emotion. A necessary (but not sufficient) condition is the requisite processing power, which I estimate at about 20 million billion calculations per sec. (we have on the order of 100 billion neurons, each with some 1,000 connections to other neurons, with each connection capable of performing about 200 calculations per sec.). As Moore's law reaches its limit and computing power no longer doubles roughly every 12 to 18 months (by my reckoning, around 2019), conventional silicon chips may not be able to deliver that kind of performance. But each time one computing technology has reached its limit, a new approach has stepped in to continue exponential growth (see "What Will Replace Silicon?" in this issue). Nanotubes, for example, which are already functioning in laboratories, could be fashioned into three-dimensional circuits made of hexagonal arrays of carbon atoms. One cubic inch of nanotube circuitry would be 1 million times more powerful than the human brain, at least in raw processing power.
More important, however, is the software of intelligence. The most compelling scenario for mastering that software is to tap into the blueprint of the best example we can get our hands on: the brain. There is no reason why we cannot reverse-engineer the human brain and copy its design. We can peer inside someone's brain today with noninvasive scanners, which are increasing their resolution with each new generation. To capture the salient neural details of the human brain, the most practical approach would be to scan it from inside. By 2030, "nanobot" technology should be available for brain scanning. Nanobots are robots that are the size of human blood cells or even smaller (see "Will Tiny Robots Build Diamonds One Atom at a Time?"). Billions of them could travel through every brain capillary and scan neural details up close. Using high-speed wireless connections, the nanobots would communicate with one another and with other computers that are compiling the brain-scan database.