Archive for September, 2009

Economics of the Robotic Revolution

Today, I was thinking about robots like I often do, and I wondered how much an individual robot would have to be in order to be economically competitive with a human being. Basically,  such a robot is cheaper than hiring a living human, then it is more advantageous for a business to buy a robot than hire a person.

A good comparison I’ve found is comparing a robot to a car. Both are highly complex machines that perform important tasks for people. In fact, in a very real sense, comparing a robot to a human is much like comparing a car to its predecessor: the horse.

Cars have many advantages over horses: they are faster, more powerful. They are cheaper and easier to buy and maintain (yes, even with $3 a gallon gas).  Cars don’t die of starvation if you ignore them for a couple weeks. Cars shield you from the elements when traveling and come with air conditioning and handy cup holders…the list can go on and on. Horses used to be the common method of getting around, but now are something of an extravagance simply because something objectively better has come along.

Robots have many of the same advantages over humans as cars do over horses. Also, in terms of economics, let’s make a few assumptions. Let’s assume that a company that buys a robot keeps that robot and uses it for labor every day for 10 years. Cars regularly last 10 years or more. Let’s also assume that the robot works an average of eighteen hours a day, using the other six to recharge, repair, and generally undergo maintenance. If the company was employing a human, how much does that much manpower cost?

The answer is quite surprising. There are roughly 3,650 days in 10 years. Times 18 hours per day, means that a company would get 65,700 man-hours of labor out of one robot. The current minimum wage in the US is $7.25, so that many man-hours of labor costs $476,325. That’s almost half a million dollars. For one robot.

Now there are also issues like the cost to power and repair those robots, but again, how much does it cost to keep a car repaired and fueled? A few hundred dollars a month? Assume $500. For 10 years that works out to only $60,000.  Still, a $400,000 robot is economically competitive. Wow. Factoring in inflation that will likely happen over the next 25 years (at which point computer power becomes high enough due to Moore’s Law), an economically competitive robot will be in the $1 million range.

Today, you can buy one human brain’s worth of processing power for much less than that. Today’s supercomputers already are smarter than humans, at least in terms of computational power. Of course, as soon as a robot that costs $400,000 hits the market it doesn’t necessarily mean that everyone will be fired and replaced by robots. Likely, the cost will have to come down to the point where the advantage becomes very obvious.