The Labor Content Fallacy
Those who argue against the risk of technological unemployment due to the coming wave of robotics, AI, and other disruptive technologies often point to what is known as the Luddite or Lump of Labor Fallacy. Their reasoning is as follows: “You Luddites believe that there is a fixed amount of labor to be done in any given economy at any point in time. When technology substitutes for labor in some industry, you think all of the displaced people must therefore become unemployed. But the reality is that when costs are lowered in that industry due to increased productivity, the savings are passed onto consumers, who spend those savings in other business sectors, which create jobs to meet the demand. Human desires are infinite, so there will always be new demand creating new jobs as automation lowers costs. Of course, there is disruption, and displaced employees will need to retrain for the new jobs, but it has always worked out this way and always will.”
Historically, this has, in fact, always been true. Outside of agriculture and large scale manufacturing, each incremental unit of a good or service (especially a service) has required significant incremental human labor. The amount of labor varies by product or service, but until very recently there was essentially nothing you could buy that would not require additional human labor hours, either in its production, distribution, sale, or delivery.
We are at the very beginning stages, however, of a new era. Those who argue the Lump of Labor Fallacy are becoming guilty of a different kind of fallacy, the Labor Content Fallacy. There is no law of economics that states that producing a good or service must require human labor. It’s just been that way so far because machines were incapable of performing the most basic of human tasks — communicating in natural language, sensing emotions, moving and operating in unstructured environments, processing information and making decisions, and manipulating wide varieties of objects large and small. Every business that wanted to innovate and deliver a valuable service or product had to hire humans, because there was simply no alternative.
In this new era, we can already see that machines and AI are steadily gaining these skills. IBM’s Watson, Google’s driverless car, Microsoft’s real-time translation, Narrative Science’s Quill, Cynthia Breazeal’s Jibo, Rethink Robotics Baxter and Sawyer, and DARPA’s Atlas are the initial versions of systems that have the potential to replace people in jobs that were historically safe from automation. Over the next few decades, as new disruptive businesses emerge employing these technologies, they will provide goods and services with minimal human labor content. The historical connection between consumption and job creation will be broken.
This phenomenon is most obvious with digital goods, which exhibit essentially zero marginal cost of production. Imagine 100 million consumers who save money due to automation in some industry buying 100 million downloads of Taylor Swift’s latest hit. How many new jobs are created by that $100 million of spending? Basically, zero. What if those 100 million people paid a dollar for a year of the WhatsApp messaging service? How many new employees would WhatsApp need to hire? Since they handled 450 million customers with a staff of less than 50, the answer is — not many.
This is a remarkable new development. Most businesses throughout history have had labor as their largest single cost and their front line employee costs scaled with their number of customers. Of course there are service businesses today where this remains true. If the 100 million people all decided to get more frequent haircuts, there is no doubt we would need a lot more stylists. The critical question is this: what direction do we think our economy and technology are headed? As more and more of our products and services are digitized and machines can handle more and more of once human-only tasks, I believe we are headed into a new kind of economy with vastly reduced labor content and therefore, far fewer jobs.
Some might argue that digital goods are too obvious an example of zero marginal labor content. What about physical goods? We will continue to crave physical objects after all, not just songs and movies and games that can be downloaded. So let’s project Amazon ten years from now, just based on the initiatives they have already undertaken. Our 100 million consumers take their savings and buy products online thru Amazon, no salesperson involved. The products themselves are made at highly automated facilities with very low marginal labor content. In ten years, Amazon warehouses will be completely automated. They are already testing prototypes of robots to replace the human “pickers” who stand for hours as a parade of Kiva robots bring shelving units to them and a computer points a laser at the item they should retrieve. Customer orders will be loaded into self-driving trucks that navigate their way into neighborhoods, where drones or ambulatory robots complete the deliveries. The trucks, drones and robots will of course themselves be built in highly automated factories. Increased demand from consumers, increased economic activity, no increase in jobs for humans.
Of course, if you extend the timeline out further, that entire process could be disrupted by future generations of 3D printing. Drop a glass on the floor and shatter it? Talk to your smartphone for a few seconds and a new one will be printed within minutes. You pay for the design (unless it was open source) and $1 per pound for the raw feed stocks. That’s it. No marginal human labor required.
What about the most labor-intensive sectors of the economy? In the US the highest growth in employment recently has been in food service, retail, education, and health care. With rising minimum wages, automation may very soon come to restaurants (Momentum Machines already has a self-contained burger making robot). Technology is poised to revolutionize education, with movements like micro-credentialing reducing the need for large faculties and administrations at big expensive institutions. Despite huge regulatory barriers, we are also starting to see innovations in health care, where within ten to twenty years we may very well be managing chronic diseases without the need for nurses and doctors.
Some will argue that the money flowing to these low or zero labor content businesses finds its way into the hands of the owners of capital or shareholders, who then invest it in new businesses — the so-called “job creators.” But if their investments are in Facebook and WhatsApp, there are meager numbers of jobs created (despite many billions in returns). Even the unusual part-time employment offered to many by Uber is slated within decades to disappear as the Uber CEO has already indicated he prefers (and has begun investing in) fleets of self-driving cars.
People who discount the possibility of technological unemployment are guilty of believing the Labor Content Fallacy. We need to begin preparing for a different kind of economy. In an age of ubiquitous smart machines, we need to shift our mindset away from the goal of “full employment”, as that will simply not be possible. While there will be infinite human wants and ever expanding amounts of work to be done, we just won’t need humans to do the work when machines can do it better and cheaper. Humans Need Not Apply.