This is the second post in a series summarising the key arguments of Ray Kurzweil’s The Singularity is near: When humans transcend biology. The first post entitled Ray Kurzweil, The Singularity and the accelerating pace of progress can be found here.
The second idea I’m going to pick out from The Singularity is Kurzweil’s prediction that by 2025 personal computers (I think costing less than $1,000 in today’s money) will have the power of the human brain.
The first component to this prediction is an assessment of the computational power of the human brain. Kurzweil looks at a number of different ways to think about this question and they all yield estimates in the range of 10(exp 14) to 10(exp 15) calculations per second (henceforth cps) – that is one hundred million trillion cps to one billion trillion cps.
Today’s personal computers, or at least those back in 2005 when The Singularity was published provided 10(exp 9) cps and an extrapolation historical increases in computing power going forward yields the prediction that personal computers will have a capacity of 10(exp 16) cps by 2025.
Kurzweil goes on to add substance to this prediction by discussing the technologies that will yield the increase in computing performance that he is predicting.
First up is Moore’s law. As many of you will know Moore’s law was coined in 1965 and described the annual doubling of the number of transistors that can be fitted onto an integrated circuit (and remember if something is doubling every year it’s rate of increase is exponential). At the time Moore predicted the trend which he had only observed from 1959-65, would continue to 1975, a prediction that was widely seen as premature. Writing in 2005 Kurzweil cited Intel’s latest estimate for the end of Moore’s law as 2020.
A common reaction to the prediction of never ending exponential growth is that it has to end some day, e.g. through physical limits. That seems intuitive to people who observe exponential increases in populations tail off over time.
Kurzweil’s answer to that, which I think is a good one, is that the rate of increase will inevitably eventually tail off within a given paradigm, but that we always invent a new paradigm that will maintain the high level trend. This means that the long term exponential increase curve is formed from a series of s-shaped curves which combine together. Each s-curve represents a distinct technology which shows slow growth in the early years, very rapid growth in the middle years before tailing off in the later years during which period the next technology is invented.
Moore’s law describes the rapid growth period of the integrated circuit era and as we are coming to the end of that period there are a number of technologies in the formative stage which are candidates for creating the next s-curve.
The first thing to note is that integrated circuits are essentially 2-dimensional arrangements and transitioning to 3-dimensions will unlock the next period of growth. Kurzweil discusses a number of emerging computing paradigms that could enable that transition:
- nanotubes and nanotube circuitry
- molecular computing
- self assembly in nanatube circuits
- biological systems emulating circuit assembly
- computing with DNA
- spintronics (computing with the spin of electrons)
- computing with light
- quantum computing
It is beyond the scope of this post to examine each of these in detail, but Kurzweil’s description of developments in each of these fields coupled with the time left in Moore’s law, and the fact that throughout history a new invention has always been found to continue the trend in increasing computing power, leave me believing it is reasonable to predict that a personal computer will have the raw power of the human brain sometime in the 2020s.
Of course having the raw power of a human brain isn’t the same as being a human brain. For that you need software, which will be the topic of tomorrow’s post.