Agatha Christie, looking back on her early life, remarked that she: I couldn’t imagine being too poor to afford servants, nor so rich as to be able to afford a car. Like most of us, I've been the other way around. I've owned a car since I was about 18, and I've never had a servant, unless you count admin assistants. Talking of cars and assistants in the same sentence, in my last tour of duty at Cadence I used to be able to say that I was so successful that even my admin assistant drove a Porsche. Unfortunately, I drove a Volkswagen. My assistant, Karen, really did race her Porsche as a hobby. Exponentials One of the biggest changes in relative values is something that we know all too well: the exponential improvement in the semiconductor industry that is another name for Moore's Law. Even those of us in this business underestimate exponential change. I remember years ago working out that at some point we would have workstations that ran at 10 MIPS, with a megabyte of memory and 100 megabytes of disk. What didn’t even occur to me was that these would not be refrigerator-sized boxes, they would be laptops. If you ran the numbers out another decade, you got even more unbelievable speeds and memory capacities, but these were not even laptops, they were smartphones. In a survey I read about a few years ago, people were asked to estimate the effect of economic growth: what would be the overall increase in national income if it grew at 5% for 25 years? Over 90% of people underestimated it, and they were so bad that only 10% were within 50% of the correct answer. Perhaps more surprisingly, the experts were not much better than the laypeople. Before reading on, think about how you would have answered. The answer is about 250% increase—the national income would be 3.5X where it started. Something that should be more widely known is the "rule of 70". If something is growing at X% per year, then it will take 70÷X years for it to double. So in the question, with a CAGR of 5%, the national income will double in 70÷5 = 14 years. So in 28 years it would double twice, namely 4X. So after 25 years a little less, 3.5X. If the timescales are run out more, the increases are even more dramatic. The world has been growing at 3.5% or so for a long time. If that carries on until the end of the century, the average per capita income will be about $200,000. That is the world average, in the US it will be in the millions. Of course, this makes some assumptions about population growth, and lack of a horrible epidemic or nuclear war that eliminates most of us. If, like most people reading this post, you live in some part of the semiconductor ecosystem, then you are used to at least some level of thinking about this exponential growth. It becomes natural to assume other industries are this way. But battery technology doesn't improve like this. It would be great if an AA battery could contain 1000 times as much power as it could back in 1990, let alone a million times as much as it held in 1970. You’d only need one for your electric car. Luckily, for mostly completely different reasons, disk drive technology has improved on an even steeper exponential, so we are not stuck with computers that run at 3GHz but with just a 10-megabyte disk. One exponential area that is stopping is CPU performance. From 1986 to 2003, processor performance improved by over 50% per year. Now the numbers are 2.5% per year, and zero might be a better estimate. This is one of the reasons for such a focus on specialized DSP processors like those from our Tensilica group, or using FPGAs for server processor offload. Baumol's Cost Disease On the other end of the scale from Moore's Law is Baumol's Cost Disease. This says that there are some areas where there has been no increase in productivity. The archetypal example from the original paper is a string quartet. It still takes four musicians for an hour, just as it did in the 17th century. However, the real wages of musicians have increased a lot since the 19th century. This is a little counter intuitive since economics links increases in wages largely to increases in productivity. But even areas of the economy where productivity has not improved still have to compete with the areas that have. If we want string quartets, we have to pay them something close to what they would get paid if they did something else. Medicine and education are two other sectors that are labor intensive, and where it still takes the same length of time to change a bandage or teach a one-hour lecture. Medicine is about 18% of the economy, and education about 8%, so together a little over a quarter of the economy is seeing negligible improvement in productivity. The "growth of the service economy" is not a recipe for growth of the entire economy. Cognitive Dissonance It is where one part of life improves exponentially, and the other does not (or worse, suffers from Baumol's Cost Disease, so is going backwards when measured in dollars) that we get the kind of cognitive dissidence that Agatha Christie commented on. Cars have not been on a Moore's Law rate of increase, but amount and quality of car we can buy for a dollar has increased many times over since Agatha Christie was a young woman in the 1910s. Servants clearly suffer from Baumol's cost disease. If you want a private butler, it will cost you whatever a top restaurant would pay him (or her? A butless? I'll just stick with butler, like actresses like to be called actors...but not so much that they want to stop the best actress Oscar). As a general rule, things involving large amounts of physical stuff like metal can't change fast. Henry Ford would be amazed by a modern car, but he'd still recognize it despite all the improvements. Alexander Graham Bell would be clueless about a smartphone. Sign up for Sunday Brunch, the weekly Breakfast Bytes email.![]()
↧