Last night I went to a great burger place in Arnold, called The Giant Burger. And I sat there waiting for my burger to arrive, I had a chance to reflect on the impact of automation.
The Giant Burger is not a fast place. It’s a place for great food. Not a place for getting great food fast. The reason is one of the employees will carefully assemble each burger to order. And because orders are big, and because she is one person, orders come out at about the rate of six per hour.
And as I was staring at her and thinking about machines, I was wondering what do machines do?
What machines do isn’t replace human beings. What they do is make less skilled workers more skilled.
Consider in the middle ages the archer. Being an archer requires a lot of skill and practice. You had to train from a young age and continuously hone your craft. In some sense, you could argue that archers were the artisans of war.
And then the gun showed up. And it wasn’t more reliable and more efficient than the original long-bow, but you could find 50 people hand them 50 fifty muskets and do almost as much damage as the archer.
In short, the gun made large armies of archers possible by reducing the skill requirement.
And that happens over and over and over again.
Look at the modern military drone. I can’t fly an F16 because I am too old, too tall and too fat. I could fly a drone. And there are more middle-aged fat guys than there are highly trained fighter pilots.
And so what happened?
We have drones all over the world killing random terrorists because we can have armies of fat people sitting in rooms flying a robot.
We have more people killed from the air than at any time since the Gulf War, and not a single pilot has done the kill.
Or look at the DaVinci system for surgery. To date, surgery was about skill. Surgeons were more athletes than scientists. With DaVinci, the skill necessary to do surgery will decline over time.
What automation does, what machines do, is they reduce the value of specialized skills and democratize those skills. And in the process make the value of the human labor declines because the number of people who can do the task increases, thereby reducing salaries.
And now software is making it worse. In the past, upgrades required new physical systems, now with software we can upgrade existing systems in place. And because of how electronics work, we can improve the intelligence of systems at the rate of 2x every 18 months.
And where it gets interesting is that in the past before software, mechanical systems had to be carefully engineered. For example, a mechanical lever has less tolerance than a computerized control system that can make micro-adjustments very quickly.
In short, we can innovate faster and cheaper than ever before in creating machines that make anybody be able to do anything.
Automation isn’t about replacing people; it’s about eliminating the need for skill and with that we remove the value of training and with that, we replace the highly trained archer with conscripts.
Which begets the obvious question:
Given that the value of skill is declining faster and faster, then that implies that the value of most human labor is decreasing, and therefore the per-unit cost of paying someone to do the job is below what people would accept.
And so when we say: Automation is killing jobs, what we are saying is that automation is causing the price we are willing to pay for humans to do jobs is decreasing.
And then we get to the policy prescriptions.
1. Some kind of universal income
One approach is to realize that there is a net surplus labor force at the current labor price, a price artificially kept high because of the minimum wage, medicare, food stamps, etc. And recognize that that group of people is going to have to die off, or leave the country for the surplus to get eliminated and in the meantime continue to extend those benefits including something like a universal income.
The problem is that that group of unemployable people is going to expand over time.
And the other problem is that there will be an increasingly shrinking set of people who will subsidize the lives of those whose skills have no value at the current price.
2. Make human labor competitive by retraining
This approach recognizes that it takes some time to build computer systems that can replace all skills and that the computer systems themselves need human operators. And so we continuously retrain people.
The challenge is that during retraining people are not employed and post-retraining the value of the labor is low. And so humans continue to experience points in time where they make less money and don’t have access to a stable income.
This also has the problem that the cost of the training has to be covered. And the folks who are making money will resent that their money is helping other people.
3. Make human labor competitive by lowering price and over time increase the price by reducing the number of people in the labor supply.
Another policy prescription is to cut those benefits such that the surplus labor becomes competitive with machines at a much lower price point, and then rely on other policies to cause the labor pool to shrink over time.
For example, a starving man will work for less than $7.25.
Cut his medical coverage, and a sick person will die off quickly.
Cut off his social security, and when he is too old to work, he will die of hunger and illness.
Restrict immigration and the number of people who enter the country will decrease over time.
The net effect will be that surplus labor will decline over time. In the short term there will be some pain, but in the long run, this will work out.
In the press, there is some discussion of the heartlessness of the tech industry because we create the machines that displace skill.
Tech is amoral. Our policy prescriptions are moral. If you are outraged with the outcomes of an amoral device, go ask yourself what policy prescriptions do you favor?