Friday, 2 February 2018

Future predictions

Here are some predictions based on current trends.

Computer chips will have one transistor per atom in 2025.

Every car will be electric by 2053.

There will be enough solar panels to cover all land on earth by 2056.

Two of these predictions are very likely to be wrong.

The first is based on Moore's law, which predicts that the number of transistors on a given size of chip will double every eighteen months.

The figure for electric cars is based on the recent increase in proportion of EVs, which in most countries is still less than one percent. The proportion may increase exponentially, and will of course stop increasing when it reaches 100%.

The figure for solar panels is based on a compound annual growth rate of 30%, which has been been happening for the past twenty years. I'm assuming that power output per area of solar panel will stay the same, which it probably won't. New panels will steadily produce more electricity for the same area, but the increase will not be large, let alone exponential.

Of these predictions, I think Moore's law is the most likely to come true. This law has held true for fifty years. I don't think atoms will necessarily stop it, since quantum computing is now a thing.

Moore's law has been enabled by the success of electronics leading to a steadily increasing budget for development of ever smaller chips. Developments have tended to compliment each other, rather than replace them. The budget is not increasing at a Moorean rate though.

These exponential growth rates are usually unsustainable since at some point they are limited by physical constraints of the real world. If things are getting smaller, of course, there is no limit. Right now there is a limit to our understanding of the very small, but if science shows us one thing it is that when we ask questions, sooner or later we find answers. The harder we look for the answers, the quicker we find them.

More interesting is Wirth's law, which states that "Software is getting slower more rapidly than hardware is getting faster." So all these improvements in the computer power are eaten up by extra complications and functionality that we don't necessarily need. I noticed this around 1992, and decided to stop spending so much time programming computers. I now wish I'd written a paper on it, like Dr. Wirth.

I'm pretty sure solar panel production will peak before we cover the whole planet, although I will not be surprised to see nature reserves clear cut for solar farms, massive floating arrays, or increased solar installation in space. They may even start making the panels up there. The economic effects of increased solar power will likely be that some electricity is effectively free, which will drive down the price of electricity, and reduce the value of the panels, making their manufacture less worthwhile. So I don't think this prediction will come true. I'm hoping to still be alive, and will be able to find out.

There will very likely be a point in the future when the only people not driving electric vehicles are stupid and rich, and I think this point will come sooner rather than later. By the time our computers are firing on subatomic logic, the majority of people will be buying new electric cars. I'm sure this will sound as ridiculous as someone predicting the wide use of steam trains in 1818, or motor cars in 1918. Also, we must not underestimate the size of the stupid and rich demographic, and its disproportionate political power. There will always be a bit of liquid fuel sloshing around, and we are unlikely to ever have 100% electric vehicles, but I think we'll be close to that long before 2053.

Here's an article from the Guardian about accelerating car sales. Here's another claiming that the electric vehicle revolution in Australia is stuck in first gear. The press is never shy to use motor-industry metaphors, but they don't realise EVs only need one gear. Also they may never have experienced the excellent acceleration of electric vehicles.