My recent Times column on Moore’s Law, technological progress
and economic growth:
The law that has changed our lives most in the
past 50 years may be about to be repealed, even though it was never
even on the statute book. I am referring to Moore’s Law, which
decrees — well, observes — that a given amount of computing power
halves in cost every two years.
Robert Colwell, the former chief architect at Intel and head of
something with a very long name in the US Government (honestly,
you’d turn the page if I spelt it out, though now I’ve taken up
even more space not telling you; maybe I will put it at the end),
made a speech recently saying that in less than a
decade, Moore’s Law will come to a halt.
The problem is that the actual electronic components imprinted
on silicon chips cannot get much smaller. They are now down to
about forty silicon atoms across and once they reach ten atoms,
weird quantum effects take over and they stop behaving predictably.
Fortunately, other experts think that may not be the end of the
story. Some other law of falling cost will come to our rescue.
Technological change has developed such inexorable momentum that it
effectively ignores wars, recessions, booms and borders, and we can
no longer switch it off if we wanted to.
I am not an economist, but as far as I can make out, for true
economic growth to happen, something somewhere has to get cheaper.
If you have to work for a shorter time to fulfil a need such as
mobile telephony, sandwiches or an airline ticket, you have a bit
of spare time left over to fulfil another need or want. That gives
somebody else a job providing for your new demand. And so on.
Sometimes things get cheaper because of different organisation of
people and things (Ryanair, Primark); sometimes because of new
inventions that cost less to make or run.
So, for example, today it costs less than half a
second of work on the average wage to earn enough to switch on a
bedside lamp for an hour. In 1950 your grandparents had to work for
eight seconds on the average wage to earn that much light. (And in
1880, 15 minutes.) Thanks to improvements in electricity
generation, light technology and productivity — which is reflected
in rising wages — you have seven and a half seconds that your
grandparents did not have in which to fulfil a different need and
provide a living to a different supplier.
In 1965 Gordon Moore, a mid-ranking Silicon Valley deity, drew a
line through just five data points and boldly deduced that the
number of transistors on a silicon chip seemed to be doubling every
18 months. His friend Carver Mead, another silicon deity, pointed
out that this made chips not only cheaper, but more reliable and
less power hungry. “By making things smaller,” Moore wrote, “everything gets better
simultaneously. There is little need for trade-offs.”
And so it proved. Decade after decade, computing costs halved
every two years. Every prediction that they would level off proved
wrong. Moore expected the limit to come when chip components
reached 250 nanometres, but they passed that in 1997 and have now
hit 22 nanometres. The inevitable, incremental, inexorable
plummeting of computing costs led to desktops, mobiles, the
internet, Twitter, better cars, better accounting, better almost
everything. It is a big part of the reason that average income has
trebled globally in real terms since 1965.
Recently another Silicon Valley guru, Ray Kurzweil, realised that Moore’s Law was at work before
silicon chips even existed. The relay, vacuum tube and single
transistor had all improved along the very same trajectory: the
amount of computing power you can buy for $100 has doubled every
two years for a century, showing no slowdown in the Great
Depression or the Second World War and no acceleration in boom
Yet predictions that computing costs would plunge even faster
now that we knew about Moore’s Law also proved wrong. In technology
you need to take each step before you can take the next. And once
you’ve taken each step, the next becomes mandatory, which is why
people throughout history have stumbled on the same inventions at
the same time. As Kevin Kelly catalogues in his book What Technology Wants, there were six
different inventors of the thermometer, three of the hypodermic
needle, four of vaccination, four of decimal fractions, five of the
electric telegraph, four of photography, three of logarithms, five
of the steamboat, six of the electric railroad. According to two
historians, no less than 23 people deserve credit for inventing the
incandescent bulb around the same time as Edison.
Such redundancy underlines the futility of trying to prevent or
even steer technological change. Opponents of GM crops and fracking
can prevent a nation sharing in the bounty of progress, but almost
certainly cannot stop the world doing so. As Kelly argues, the
technium (the sum of our technologies) marches onward, selecting
its inventors rather than vice versa. It is almost as if it is
With Moore’s Law, there are plenty of possible ways to keep
costs falling even if component size stops shrinking. Some are
mundane — better run companies, cheaper materials. But others are
high-tech. If transistors can communicate using light rather than
wires, speeds could jump, energy waste fall and costs plunge again.
Or if last week’s Nature magazine is to be
believed, carbon nanotubes — essentially folded versions of the
sexy new substance called graphene — could one day “take us at
least an order of magnitude in performance beyond where you can
project silicon could take us,” according to Philip Wong, of Stanford
Moore’s Law is a reminder that by far the most useful thing you
can do for humanity is to lower the cost of something. Yet far too
few people in government, charities, companies, churches or aid
organisations think of this as their duty. More’s the pity. Let’s
hope the repeal of Moore’s Law is followed by the enactment of many
more such Moore’s laws.
Mr Colwell’s job, by the way, is director of the Microsystems
Technology Office at the US Federal Government’s Department of
Defence’s Defence Advanced Research Projects Agency (Darpa).