This is a companion piece to my 2020 essay “Creative” jobs won’t save human workers from machines or themselves, so I recommend rereading it now. In the three years since, machines have gotten sharply better at “creative” and “artistic” tasks like generating images and even short stories from simple prompts. Video synthesis is the next domino poised to fall. These advancements don’t even represent the pinnacle of what machines could theoretically achieve, and as such they’ve called into question the viability of many types of human jobs. Contrary to what the vast majority of futurists and average people predicted, jobs involving artistry and creativity seem more ripe for automation than those centered around manual labor. Myth busted.
Another myth I’d like to address is that machines will never render human workers obsolete since “new jobs that only humans can do will keep being created.” This claim is usually raised during discussions about technological unemployment, and its proponents point out that it has reliably held true for centuries now, and each scare over a new type of machine rendering humans permanently jobless has evaporated. For example, the invention of the automobile didn’t put farriers out of work forever, it just moved them to working in car tire shops.
The first problem with the claim that we’ll keep escaping machines by moving up the skill ladder to future jobs is that, like any other observed trend, there’s no reason to assume it will continue forever. In any type of system, whether we’re talking about an ecosystem or a stock market, it’s common for trends to hold steady for long periods before suddenly changing, perhaps due to some unforeseen factor. Past performance isn’t always an indicator of future performance.
The second problem with the claim is that, even if the trend continues, people might not want to do the jobs that become available to them in the future. Let me use a game as an analogy.
“Geoguessr” is an e-sport where players are shown a series of Google Street View images of an unknown place and have to guess where it is by marking a spot on a world map. The player who guesses the shortest distance from the actual location wins. Some people are shockingly good at it. Some tournaments offer $50,000 to the winner.
Anyway, some guys built a machine that can beat the best human at it.
This is a good model of how technological unemployment could play out in the future. Geoguessr, which could be thought of as a new job that was made possible by advances in technology (e.g. – Google Street View, widespread internet access) was created in 2013. Humans reigned supreme at it for 10 years until a machine was invented that could do it better. In other words, this occupation blinked in and out of existence in the space of 10 years.
That’s enough time for an average person to get trained and to perform it well enough to become an expert and net a steady income. However, as computers improve, they’ll be able to learn new tasks faster. The humans who played Geoguessr full-time will jump to some new kind of invented job made possible by a newer technology like VR. There, humans will reign supreme for, say, eight years before machines can do it better.
The third type of invented job will exist thanks to another future technology like wearable brain scanners. The human cohort will then switch to doing that for a living, but machines will learn to do it better after only six years.
Eventually, the intervals between the creation and obsolescence of jobs will get so short that it won’t be worth it for humans to even try anymore. By the time they’re finished training for it, they might have a handful of years of employment ahead of them before being replaced by another machine. The velocity of this process will make people drop out of the labor market in steadily growing numbers through a combination of hopelessness and rational economic calculation (especially if they can just get on welfare permanently). I call this phenomenon “automation escape velocity,” whereby machines get faster at learning new work tasks than humans, or so fast that humans have too small an advantage to really capitalize on.
This is a scenario shows how the belief that “Machines will never take away all human jobs because new jobs that only humans can do will keep being created” could hold true, but at the same time fail to prevent mass unemployment. Yes, humans will technically remain able to climb the skill ladder to newly created jobs that machines can’t do yet, but the speed at which humans will need to keep climbing to stay above the machines below them will get so fast that most humans will fall off. A minority of gifted people who excel at learning new things and enjoy challenges will have careers, but the vast majority of humans aren’t like that.