On Wednesday I heard Ray Kurzweil speak.
Kurzweil gave an abbreviated version of the gospel he’s been preaching for the last twenty years: that technology evolves on exponential curves, whereas intelligence has evolved to make linear predictions. “If you can measure the underlying properties of an information technology,” he said, “they follow smooth exponential arcs.” This is a pattern that has held through the Great Depression, both World Wars and every recession of the last century.
When people talk about a technological growth curve plateauing, they’re looking at one factor: speed, energy, cost, etc. People have observed Moore’s Law plateauing of late. But a slackening in one curve doesn’t mean progress stops on all others. Kurzweil gave the example of the MIT lab computer he rented time on when he was an undergrad versus the iPhone he had in his pocket: a billion-fold price/performance increase in 30 years. And the chart of MIPS per Dollar grows at a smooth exponential arc through the past century.
The reason technology grows so exponentially, Kurzweil explained, is that it’s an evolutionary process. The first computers had to be designed on graph paper; now they’re designed by other computers. This generation’s technology helps to make the next generation’s. As a result, the rate of technology adoption accelerates as well. It took several decades for phones to be universally adopted, only a couple decades for phones and only a few years for cell phones. Smartphones have taken off even faster.
Of course, it wouldn’t be a Kurzweil talk without some crazy sounding predictions. I copied down the following:
- Drugs that can suppress the fat insulin receptor gene in humans, keeping our bodies from stockpiling calories (an evolutionary holdover);
- Solar energy has doubled in productivity every two years; at that rate, solar power could meet 100% of the human race’s needs in 20 years (granted, those quoted doubling figures are talking about photovoltaic cells, and most engineers agree that further growth in efficiency would require some developments in nanotechnology – that being said, it’s not impossible);
- By 2029, $1000 in computation will have 1000x the capacity of the human brain (this was a blurb on a slide that he didn’t dwell on; did he mean speed? storage space? both?);
- Moore’s Law refers to the processing power of integrated circuits, what Kurzweil calls the “fifth paradigm” in information technology (vacuum tubes being the fourth). The sixth paradigm – a three-dimensional circuit substrate – could be possible by 2020.
I don’t know that I like the flavor of his Kool-Aid yet. But his predictions on information technology have borne out in the past. And – what’s most important to me – his reasoning makes sense.