Robots and Jobs

In this section I'd like to talk about robots and jobs. Whenever I go and give a public talk about robotics, somebody in the audience always ask me about robots and jobs. Aren’t robots going to take jobs away from people? And it's true, robots will take jobs away from people. But there are jobs that I believe people shouldn't do. And that's because those jobs are dangerous.

So if we look at the data shown here, we look at the number of fatalities per 100,000 workers in a number of different occupational areas recorded at two points in time. And we can see that there are some industries which are really quite dangerous. People can go to work in one of these industries and not come home at night because they've been killed on the job. And to me that's non-acceptable. If there are dangerous jobs, there should be machines doing the work instead of humans. 

I'm going to focus a bit on the mining industry because that's an area where I spent more than 10 years developing various robotic technologies for underground and surface mining. Let's first of all look at some fatality data from the United States mining industry compiled over a period of about 100 years. 

The line graph shows the fatality rate normalized to production. Then we see a very happy reduction in that fatality rate. The black line is for the coal industry and the red line is for  non-coal industry which is generally surface mining operations. They tend to be safer than underground coal mining.

So while there has been a very considerable improvement over time. We observe just as we did for road accident fatalities that the rate of improvement is plateauing and it's not plateauing to zero. It's plateauing to a value above zero. So there is some residual rate of fatalities that will occur in mining operations. Despite 100 years of continuous improvement, we are really struggling to get this rate down to zero.

If we look at the types of accidents that occur in mines today, we see that the causes are really quite varied. Some of them are unfortunate encounters between human beings and machines, conveyor belts and trucks and so on. Some of them are electrocutions, people slipping and falling, relatively few of them involved people being killed by falling rocks, which is perhaps what most people would consider to be leading cause of death in an underground mine.
 
So what can we do about this? Well the important thing is to remove the human beings from hazardous situations. Here is a type of vehicle called a load haul dump unit, or a scoop train if you come from Canada. And it's used to carry ore in an underground mine. Now once upon a time these machines were all driven by a driver who sat here. The problem is that these machines go in to dangerous parts of the mine. So what some mining companies do now is to remove the operator to a safe, remote location. Here the operator is able to control the machine with the joystick and she can see the status of the machine through a view of the camera and the position of the vehicle superimposed on a map of the mine. And that situational awareness information is coming from various sensors fitted to the machine.

So what we're seeing now is a separation between the person doing the work and the place where the work is occurring. And this happens in many other situations as well. An underwater vehicle which is doing some task deep beneath the ocean. The operator is at some comfortable work station on board a ship. Consider the case of a Mars rover. It's doing the work of geologist and chemist on the surface of Mars whereas the actual geologist and chemist are sitting at work stations on earth. 

Consider the case of robotic surgery that we look at earlier. The surgeons are doing their work at work stations on the left hand side of the picture where the actual work, the actual surgery is occurring inside of the body of the patient. And that's being done by the robots here. Now the robots are connected to the work stations by cables. So there's no reason in fact that those cables couldn't be longer. They could in fact be an intercontinental telecommunications link. And this has actually been demonstrated. We've had surgeons on one continent performing an operation on a patient in another continent. So the worker and the place of work are now quite distinct.

So traditionally, the worker has been at the work place. The worker has been on the machine or controlling the machine directly. Right now with remote control technology, it's possible that the worker can be remote from the machine. But in the future I think what's going to happen is that the machines will be smarter and they'll be able to do a lot of the work all by themselves. So we will move to a situation where one worker will be remotely managing many machines. That means that the worker is not joysticking the machine in the traditional sense, they’re giving a high level command to the machine. They’re saying machine A go off and do this task. Machine B go off and do this task. Machine C go and do this task. And if one of the machines has encountered a problem then it would come back to the operator. Machine B may come back to the operator and say, "I don't know how to do this thing that you've asked me to do." So the operator can give machine B some more detailed instructions.

So it's like the way a human manager manages a number of human workers. But now we have a human manager managing a number of intelligent machines. Perhaps sometime in the medium future, the machines will be smart enough not to need human managers. But right now we need a combination of human intuition, intelligence and experience combined with the relentless work ethic, the tirelessness and the reliability of robot workers. 

So in the future I believe work will be a partnership between human managers and robot workers. And what this enables is the same or greater level of productivity perhaps with lower operator skill levels, because the skill is going to be encapsulated in the robot worker. The robot worker is going to do the detailed skillful work and the human manager is going to tell the robot where it should be deploying its particular skill. Machines are going to suffer much less damage because they're going to be driven by computers rather than by human beings. The computer is going to ensure that the machines are driven safely and within their operational envelopes.

And finally we'll be able to achieve force multiplication. That means that more work will be able to be done by fewer people because they're going to be assisted by robots. In order to achieve this, we're going to need high communications bandwidth between where the robots are doing their work and where the human managers are sitting at their work stations. We're going to need the machines to have some level of local autonomy, some level of local automation. We're going to have to take a systems approach to the way we do tasks; where do we situate the people, where do we situate the robot, how much intelligence should we embed in the robot, how much communications do we need. So it's going to be a new way of thinking about work places and how humans and robots interact.