Wednesday, May 29, 2013

Who Will Prevail – Robots or Supermen?

Who Will Prevail – Robots or Supermen?

We are currently making great progress in artificial intelligence and automation. Moore's Law is the rule that the number of components that can be placed on a circuit board doubles about every two years. Assuming that Moore's Law holds true in the next few decades, we can imagine that by some time within the next hundred years robots may have intelligence equal to or greater than the current intelligence of human beings.

Does that mean that robots are likely to take over the planet within 100 years? Not at all. There are two reasons why it is rather unlikely that robots will become the rulers of the planet within the next 100 years.

The first reason is that it is quite possible that humans themselves may become smarter within the next fifty years or so. We are learning how to tinker with our own genes. Before long, we may discover genes that can be added or tweaked to give human beings more intelligence. So if we imagine very brainy robots inhabiting our planet at the end of the century, we can imagine that their rivals will not be people as intelligent as us, but perhaps people much smarter than we are.

So let us imagine it is the year 2100. The world is filled with millions or billions of extremely smart robots. But there are also many people in the world with IQ's higher than 200. Which is likely to get the upper hand – the robots or the supermen? 
 
Some would argue that robots would have the upper hand, because humans have only a little extra room to store additional brain cells, but it is easy to make a robot that has twice as much of a CPU, or five times as much, or ten or twenty times as much. Imagine you were a master genetic engineer who can tinker with the arrangement of the human body. You couldn't simply design human beings with heads twice as big, because the head has to fit through the narrow slot of a woman's birth canal when a baby is born. So it seems we can't design humans that are, say, ten times smarter, unless we completely redesign the human body, which might be a nightmare.

For robots, there is no such limitation. A robot can be almost any size or shape. Even if a humanoid appearance is chosen, the chest of the humanoid robot can be used to store additional CPU units. A robot can also have a non-humanoid appearance. If the robot looks something a battle tank, there will be abundant room to store large CPU units. So we can imagine robots with thinking units weighing 100 times more than a man's brain. That would seem to give robots the upper hand.

But humans do have two very substantial advantages which may assure their predominance over robots for at least a century. The first advantage is that it is easier for humans to move around, without being tethered to any power cord, and without requiring any battery replacement or battery recharging. Imagine an extremely intelligent robot of the year 2100. How would such a robot get around? Today almost all robots need to be plugged in to an electrical source. Is there any way that a robot could be able to move around independently as a human does?

We can imagine solar-powered robots like the Curiosity robot landed on Mars. But solar power can only provide limited power for any robot that is the size of a car or smaller, and the Curiosity robot is a slow-moving sluggish thing compared to a human being. Solar-powered robots may be quite useful, but are unlikely to be mobile enough to gain dominance over the human race.

We can also imagine battery powered robots which recharge themselves using the electrical grid. But that would probably not be sufficient to give these robots the power to take over in a world in which humans control the electrical grid.

Here is another reason why it would seem very difficult for robots to take over the planet within a century: for a long time it is going to be very hard for robots to self-reproduce (but it is very easy for humans to reproduce). An advanced robot is likely to be made from components that are manufactured in many places around the world through highly specialized tools and manufacturing processes. It is hard to imagine a robot within the next 100 years having the ability to make a copy of itself from metals and raw materials that can be scavenged outside of a factory. If robots are not able to independently self-reproduce in the next 100 years, they are unlikely on their own to duplicate themselves to reach a population level at which they can dominate us. We humans will still be in charge of how many of them there are. By reducing their growth, we can presumably keep our position as the lords of the planet, for quite a few decades to come.

However, centuries from now this situation may change. Robots may well acquire all the mobility advantages of humans, and may well gain the power to reproduce themselves even faster than people can reproduce themselves.

If that ever happens, our days as lords of the planet may be over. 

No comments:

Post a Comment