THE SINGULARITY….And now for something completely different. Really, really different.
A couple of weeks ago the nice folks at Kurzweil Technologies sent me a copy of Ray Kurzweil’s The Singularity Is Near. Kurzweil’s thesis is pretty simple: he believes that the only thing standing in the way of genuine artificial intelligence is computing power. Right now, a $1,000 PC has enough computing power to emulate the brain of an insect, but once we increase that by about a billion times or so, we’ll have a $1,000 computer with the power of a human brain. Given current doubling trends, that should happen in about 20 years. (You can read a more detailed version of this argument here.)
But there’s more: once we have true intelligent machines, these machines will themselves create the next generation of machines. However, being machines, they’ll do it way faster than we could. So we’ll go from machines with the intelligence of a human to machines a million times smarter than humans in a very short time. This will shortly bring us to the Singularity, a point representing a “profound and disruptive transformation in human capability,” which Kurzweil estimates will happen in 2045.
Now, it so happens that I didn’t really need to read Kurzweil’s book to learn this stuff, because I already think he’s right. The basic hardware and software trends seem pretty indisputable to me, and the only serious arguments I’ve ever heard against the eventual development of genuinely intelligent machines all boil down to a thinly veiled belief that there just has to be something more to human intelligence than mere neurons and biochemistry. Well, no there doesn’t. The pope’s opinions notwithstanding, the evidence to date suggests that the brain really is just a biological computing device.
So that’s that ? though feel free to argue the other side in comments. With that said, however, it turns out that I do have a bone to pick with Kurzweil over one of the trend charts that litter his book. Basically, he argues that the pace of change has been accelerating over time, so that major inventions are being created ever faster as time goes by. 10,000 years ago it took several thousand years between major inventions (agriculture –> wheel), while a century ago it took only a few decades (telephone –> computer).
Fine. But his cleverly constructed chart cheats: it stops about 30 years ago. So I decided to extend it. My version of his chart extends to last month (see pink shaded area), and it indicates that major, paradigm-busting inventions should be spaced about a week apart these days.
(Sorry for the lousy quality of the chart. It’s a scan from the book. Futurist though Kurzweil may be, his publisher has apparently declined to join Amazon’s “Search Inside” program, which would have allowed me to capture a clean copy of the chart right off the screen. He needs to have a word with them about this.)
So what gives? Seems to me that the Singularity should be right on our doorstep, not 40 years away. And while 40 years may not seem like all that much in the great scheme of things, it means a lot if you’re 46 years old. Which I am.
So what happened?
UPDATE: Via email, Ray Kurzweil responds:
It isn?t valid to extend a log-log plot. A progression is valid by showing exponential growth along a linear time axis, so a graph with a linear x (time) axis and a log y axis can be validly extended (provided of course that one has analyzed the paradigm being measured and shown that it will not saturate to an asymptote). So that is why I analyzed extensively the limits of matter and energy to support computation and communication, as well as the specific technologies that could support these densities including analysis of the heat and energy issues. And an exponential trend (a straight line on a plot with a linear time axis and a log y axis) or a double exponential (an exponential on a plot with linear time axis and a log y axis) does not reach a mathematical singularity but it does reach fantastic levels eventually.
So the point of the log-log plot is simply to show that a phenomenon has in fact accelerated in the past. It is not valid to extend the line. For one thing the log-log plot cannot go into the future because that is the nature of the log time axis. If one wanted to extend this trend, one should plot it on a linear time (x) axis showing exponential progression of the paradigm shift rate. I did that in another chart where I show the adoption times (for mass use) of communication technologies such as television, telephone up to cell phones, etc.