Computers are a wonderful invention, of course—I’m typing this on one, and they make the entire concept of worldwide blogging possible. The struggle is that as frequently as they create jobs, they can destroy them by automating work that had previously been done by humans.
It’s a common misconception that automation is something that can only happen to blue-collar, manual-labor jobs. This misconception arises because factory jobs were among the first to be automated, and they were the first to be automated because the nature of the assembly line makes it easy for assembly-line jobs to be automated.
In general, the less context-driven the task, the more easily automatable it is. To construct, for instance, a paint can on an assembly line, you do the same tasks in the same way in the same place each time. All you need is the primitive robotics. Nothing changes, and in the off-chance something does change, it’s mostly just a matter of a one-time retooling of the factory and reprogramming of the computers.
Where this falls short of the mark is in highly context-driven tasks, where what needs to be done is highly variable in more ways than a computer programmer can easily deal with and/or where particular nuances are important. The prime example of this is in customer service. While the automated phone response systems (“Press 1 to _____”) have been around for a while, ultimately their goal is to transfer you to the appropriate department and give the actual human who will eventually take your call a bit of a heads-up on why you might be calling.
Janitorial services are a blue-collar task that by and large has not been automated. Why? Largely because it’s much more difficult to automate it. Each building a janitor might need to clean is different, each mess is different, requirements may vary, impediments may vary, and so forth. In order to automate such a process, a computer would have to be programmed in such a way to account for and adapt to all of these variations without manual reprogramming.
However, as computing and robotics technology improves, the number of automatable jobs that exist increases. Jobs that had previously been thought to be human-only, such as retail checkout and certain types of legal research, can be outsourced to computers (though retail cashiering may not, in fact, die out completely as it depends in part on consumer reaction).
No matter how you cut it, a computerized system is almost always going to be less expensive than a human employee in an easily-automated role. Even if you allow 12 hours a week of shutdown for things like updates, that’s still 156 hours a week of uptime—computers don’t need to “sleep” as often.
Computers also require less money. Certainly, they require electricity to operate like humans require food and water, but on a sheer energy intake basis, computer electricity is much cheaper. A Big Mac from McDonald’s, for instance, costs about $4.33 and contains 550 food calories of energy (equivalent to 639 watt-hours). This is equivalent to $6.78/kWh, and I deliberately chose a cheap and unhealthy food to minimize this ratio. Electricity, by contrast, typically costs less than $0.50/kWh – sometimes more, sometimes much less, depending on where you are, local taxes, available sources, and so forth. But it certainly costs less than $6.78 per kilowatt-hour.
The laptop I’m typing this on pulls 90 watts of energy—in other words, in one hour it uses 90 watt-hours. If I needed the processing power of a hundred of these laptops to run my factory or business or whatever, I’d pull 9 kilowatts, and if I ran them for 156 hours a week I would use 1404 kilowatt-hours of energy per week. Even at 50 cents per kilowatt hour, this would be $702 per week or $36,504 per year. Even if we say that the cost of cooling, maintenance, software, programming, etc. raises the cost to $120,000 per year, that’s the equivalent cost of about three union employees to do the work of thirty per shift.
Not to mention computers don’t care about retirement programs, health insurance, facility safety, time off, worker’s compensation, collective bargaining… the list goes on. Automating also means you reduce the number of needed employees, which may reduce the cost of certain regulatory compliances. When you take this into consideration, it’s easy to see why many companies that can automate will.
Simply stated, for easily-automatable positions, high-wage humans can’t compete with robots. But automation is not, of course, the only headwind facing the job market... but what is is a subject for a later post.
