What happens if Moore’s Law continues to apply for the next sixty years? Strange things, it would appear.
I should warn you: today I’m dealing with a subject that is really rather theoretical, but one which I’ve been wrestling with for some time. Today’s graph perfectly illustrates the problem, in my eyes. Unfortunately for me, I plucked it from a site that only contained the graph, with little in the way of text and explanation, for I was actually quite curious what the person who posted it was trying to say to us.
The Six Million Dollar Man
So what’s the problem? The easiest way to explain this is probably to tell you how I came across the problem myself. Basically I heard that the series ‘The Six Million Dollar Man’ was to be aired again. For those that don’t know, The Six Million Dollar Man was a series in the 70s about a pilot who, after a terrible crash, was rebuilt using all sorts of technological (‘bionic’) gadgets. A laser eye, two super legs and a super-strong arm, and all that for the pretty price of – yeah, you guessed it – six million dollars. This half superman was then able to spend many happy episodes winning sprints against cars, reading secret reports at 500 meters and catching crooks.
Enter the remake. The question was of course: so what are they going to call the successor to this series, exactly? To correspond with the seventies, you should of course have the feeling that the bionic man cost an arm and a leg (no pun intended!), so the filmmakers’ answer was simple: The Six Billion Dollar Man. Sounds logical, but it provoked the makers of NPR’s Planet Money podcast to then ask the question as to whether it was indeed so logical that this man had become 1000x more expensive over the last 40 years. The conclusion was that, looking purely at the technology, The Six Million Dollar Man should actually be called The Twelve Thousand Dollar Man, since the price of bionic laser legs has dropped hugely over the intervening 40 years. Not that there are such things as bionic laser legs, but assuming the price development of technological products, that would be the price you would expect. Other sites too broke it down nicely, and the outcome was basically the same: The Six Million Dollar Man would be dirt cheap these days. You would reach a price of 30 million only by adding the cost of medical staff and the costs of the operation. The message? Technology is much cheaper, doctors are much more expensive.
What I’m illustrating here, using the example of The Six Million Dollar Man, is a phenomenon that is of course described by Moore’s Law, the ‘law’ observing that the overall processing power of computers will double every two years. What’s more, that doubling also corresponds with a strong price decline, and that’s what we’re dealing with here. If you wanted to buy a computer ten years ago with the same memory, speed or user-friendliness of a new one today, you would probably have to have paid ten times as much. I remember buying a 200MB external hard drive for 200 Dutch guilders: a USB stick with 32 gigabyte is yours today for just 10 euros. 64 times the memory space for 1/14 of the price: and, you guessed it, that price effect is also reflected in the aggregated level of an economy.
That’s precisely what the graph shows us. Primarily it’s referring to the orange line, which shows the nominal share of the processing industry in US GDP. What we see is that this share has declined gradually over the last 60 years. This will partly be due to the services sector becoming increasingly important, but is also partly down to the price effect mentioned above. It becomes clearer if we look at the blue line, which shows the real importance of the processing industry, which has actually remained relatively stable over this period. In other words, if we take account of the price effect, the importance of the processing industry has remained more or less constant over the last 60 years. However, since the prices of the products have declined in relative terms over that period, the share in the economy has become smaller.
And that’s where we hit the problem. Suppose this trend were to continue over the next 60 years? Then one day, a computer will cost just one euro, and a capable robot will be yours for 50 euros. We will of course then buy more computers and robots, but how many robots can you actually fit in your home? In terms of sales (price times volume), the volume will increase, but if the price continues to decline, the orange line will also continue its path downwards. Suppose the orange line drops even more, to 5%, to 1%, to 0.5%. In real terms (adjusted for the price effect), the blue line would easily remain stable, or even rise. You would then get the absurd situation where something which is of marginal import for the economy as a whole in sales terms, is still vitally important in real terms. And where the few people who still work in that part of the economy (the rest have been automated into redundancy) are still very important according to the GDP statistics. I can already imagine the official statistics press release: “real growth in the second quarter 2% negative, as Bob had surgery on his appendix.”
I’m being flippant, of course, but still. It does beg the following question: suppose that the concept of a GDP would not be invented until 2070. Which weight would the processing industry then get in the nominal economy? 0.5% it seems to me. And nominal? Again 0.5%, in my eyes – there is no reason to say 20% at that moment, as you’d have to have a historical context to arrive at that number. If this is correct, it means that your starting/calibration point would have a significant impact on the further development of your real growth.
I realize that it’s pretty unlikely that the orange line would ever fall to 1%, let alone even lower. Which is why there are plenty of people who now suggest that Moore’s Law has had its day. However, as (relative) demand for products increases, the (relative) price declines. What’s more, chips are not by any means the only factor in the production cost of computers. The price of metals, casings and wiring also play an important role, of course. The point I’m trying to make is that everyone can see that the extreme situation ends in illogical results. The question is, might this already be the case, but nobody’s realized it yet as the change has happened so gradually. And it’s also bound to be the case that the statistics are re-calibrated with some regularity to prevent further distortion. However, looking at the graph, that doesn’t seem to have prevented the stated trend occurring.
So that’s the long and short of it. Anyone know the answer?