Despite fear mongering, AI isn’t going to kill us all, but in the near future it will cause a lot of societal disruption for which we are ill prepared.
Growing up a Star Trek fan I believe we can create a better society tomorrow. I subscribe to Pinker’s Better Angels theory, despite some occasional backslides encountered on the road of history. As such, I see AI bringing a better future-–in the long run. In the short term, it’s going to be a bumpy ride.
Star Trek itself didn’t do much with AI, other than Data who was a one-of-a-kind android. The rest of Hollywood, on the other hand, has given us a variety of potential AI nightmare scenarios. There are robot uprisings on a per capita basis (Blade Runner, Westworld), a global AI against humanity as a whole (Terminator, The Matrix, Avengers: Age of Ultron), AI in the military (Robocop, Minority Report–albeit not quite traditional AI), and one-off AI chaos (2001: A Space Odyssey). There are also friendlier, or at least less hostile, outcomes ranging from AI outgrowing us (Her), to being a new, friendly type of entity (Short Circuit). Classic sci-fi stories from Asimov often get more nuanced than Hollywood in the implications.
AGI (Artificial General Intelligence) also known as strong AI, which is what people see in movies of AI, that can think across a wide variety of tasks, is a long way off. A computer can beat us in chess or detect cancer on an image better than a human, but that software can’t then expound on who is cuter, WALL-E or Number 5. It’s similar to how the robots that can assemble cars or paint a Rembrant are unable to take the first step in rock climbing. It’s not just a lack of legs, it’s a lack of knowledge how to use legs. They are purpose built and not general. This is known as weak AI or narrow AI, it's where we are today.
We may one day have to fight robots for the survival of humanity, just like we may one day drive flying cars and live on the moon. I use these references because they were first promised to us some 70 years ago but have not yet come to pass and may not in my lifetime. Roy Amara famously said, “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” (Consider how computers today compare to the computers on the original Star Trek series.) So, while an AIpocalypse may come to pass, it’s further out than we think.
Hollywood . . . has given us a variety of potential AI nightmare scenarios.
More concerning is the impact of AI on society. In part one of a two-part article, we’ll look at the impact on labor. In part two we’ll look at the new means of information construction and dissemination, and its impact on society.
I’ve written and spoken prior about technology disrupting labor. We have seen time and again technology destroy entire lines of work. This isn’t a bad thing. Milk men, elevator operators, and toll booth collectors all went away (or were significantly reduced) due to technological improvements and most people feel we are better off for it-–just not those who worked in those jobs and found themselves without options.
As we look at past transitions, we see one of two things held: it took a number of years and/or affected a relatively small number of people. In 1800 83% of the US workforce was in agriculture but by 1900 this number dropped to roughly 30% (that’s technically the number of people living on farms, not working on farms, but the two have a close correlation). While significant in scope, this change happened over a century, or roughly 4-5 generations. In the case of elevator operators, we can see it peaked at 120,000 in the US in 1950 and dropped by half a decade later. While that’s quite rapid compared to the slower decline in farming, the US population in 1950 was 150 million, meaning those 60,000 lost jobs were 0.04% of the labor market. To put that in perspective, the last week for which data is available at the time of this writing, June 10th, 2023, the Department of Labor listed 251,000 initial unemployment claims (not seasonally adjusted). Out of a population of roughly 332 million this is 0.075% of the population. In other words, that 0.04% change over a decade is half of what we see in a typical week during a strong labor market.
This matters because small pockets of displaced labor (a specific field) can be absorbed by the economy as a whole. Likewise larger changes, such as what happened to the farmers, can be absorbed over time.
“We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” --Roy Amara
This is not to say it always happens smoothly. The Luddites in England protested the introduction of the power loom and destroyed the new machines followed by the Swing Riots a few years later. The latter part of the nineteenth century in the US saw significant unrest and protests from the farmers who saw their prosperity and political influence weaken. We saw similar frustration with factory workers towards the end of the twentieth century and start of the twenty-first century as US manufacturing declined. While cunning, albeit deceitful, politicians tapped into this frustration blaming offshoring for the loss of jobs, the truth is 87% of the jobs lost were due to technology, not geopolitics. (Note: The study was only of manufacturing job losses 2000-2010, although it's reasonable to assume there was not a magic manufacturing technology introduced in 1999, but rather a progression of improvements from the mid-century onward.)
In the coming years, long before we get enslaved or hunted down by robots (if ever), we will have people displaced in their jobs like never before. The number of jobs and speed at which they’re lost will be unprecedented.
The closest analogy would be post-war, where infrastructure has been destroyed and the soldiers returning home represent an influx of unemployed labor. When that has happened in the past two things were true that don’t necessarily hold today. First, the damaged infrastructure needed to be replaced (or built up in the case of the US suburbs post WWII). Factories, buildings, houses, roads, all needed to be built, not to mention supporting infrastructure like stores, furniture, and supply chains. Today our infrastructure is generally intact. (Side note: in the US infrastructure gets a C-, and we need to reimplement supply chains due to geopolitical changes, but we don’t have to rebuild from the ground up as in many post war countries.) Second, the labor needed at those times was relatively low skilled. Mid-century construction jobs weren’t high-tech. Even welding was a skill that could be learned in weeks.
According to The Future of Jobs Report in 2020 from the World Economic Forum by 2025 85 million jobs may be displaced globally. The report also noted that 97 million new roles may emerge from the introduction of technology to new areas. This is consistent with history; technology creates job opportunities as well as job losses.
The catch, as implied above, is that there is a time lag. As steam powered farming machines replaced workers on the farms, new factory workers were needed to build the farming machines, plus the machines that made the farming machines, and other machines transported the farming machines, and the roads they were transported on, which in turn needed their own machines to make the roads, etc. As that happened over decades the labor force had time to adjust. This will not be the case with the disruption from AI.
Milk men, elevator operators, and toll booth collectors all went away (or were significantly reduced) due to technological improvements and most people feel we are better off for it.
The world will lose tens of millions of jobs in the coming years. Any single country could be losing millions to tens of millions of jobs alone. It may well be years before many of the new jobs arise. Even if I’m wrong on the speed of arrival of the new jobs, and I may be, people will not be trained for those new jobs. This is the key. A farmer in 1870 could move into a factory job in 1870 because it was relatively unskilled labor, making the cost of transition fairly frictionless. That does not hold for today’s changes.
The result of all this will be significant labor market liquidity issues, economic volatility, and societal unrest. Unemployment will tick up by whole percentage points and stay there until we get to the other side of the shift. That’s on the first order; unscrupulous politicians will no doubt leverage it in ways that create secondary problems due to shortsighted scapegoating as we saw with the misplaced protectionist approaches taken as offshoring was blamed for job losses caused by technological improvements a few decades back.
Before you despair, there are options. We can, and will, retrain the workforce. The nature of this retraining is the future of education. I’ve long argued that our current system is misaligned. We educate people for 10-20 years, K-12 and optionally college, and then expect them to work for decades to come with no additional training. This is not realistic in the twenty-first century. Instead, we need to recognize that there will be periods of retraining throughout someone's career. This is ideally some number of hours of training per year, but then also likely periods of more intensive training, on the order of months, roughly every 5-10 years. It’s not college, not even necessarily a one-year master’s degree, but it’s likely scores to hundreds of hours of training over some number of months. We don’t have this type of training available on a wide scale yet, but we will need it.
In WWII the US Department of War (predecessor to the Department of Defense) created the War Manpower Commission. It was designed to reorient US labor to wartime needs. While we don’t need anything quite so heavy handed as what was needed during WWII it does set a precedent for labor guidance by the government.
Government is needed because, as with the military, postal service, and other core government services, no private corporation has the incentives or capabilities to do it. To be clear, existing educational organizations (e.g., private schools, public colleges, online platforms) can provide the actual training. What is needed is guidance and support.
Guidance will be in the form of giving direction. The free market can figure out what training is needed, and what certificates are valid versus garbage diploma mills. However, on its own that could take a few years and the economy can’t have tens of millions of employed people waiting around for that to be figured out. Government, guided itself by industry and academic experts, can accelerate this.
In the coming years, long before we get enslaved or hunted down by robots (if ever), we will have people displaced in their jobs like never before. The number of jobs and speed at which they’re lost will be unprecedented.
While the classes themselves can be done with existing infrastructure, incentives and support are needed in a way that only the government can provide at scale. Federal grants and loans are used to help people go to college. Similar programs can be implemented for the non-college training programs that will be needed in the coming years. Equally, if not more important, will be other economic incentives. For example, companies laying off employees may receive incentives if, as part of the unemployment, they help employees find and enroll in such retraining programs. The laws could be changed to allow people to collect unemployment while in such training programs, even if not available for work. Other programs could provide food or housing support while people enroll in such retraining.
In the US the Democrats have traditionally supported these types of government programs while Republicans have not. That misses an opportunity. The Brookings Institution notes how Democratic districts are faring better economically than Republican districts. The AI revolution represents a chance to reset the playing field and bring more of the economically disadvantaged (from any district) into the new economy. Because these new jobs are often online (the ones created by AI, there will be a different set of physical jobs created by the manufacturing reshoring activities in the coming years), they will be available to anyone in the country, regardless of location.
With government and societal support, we can reduce the time this transition in the labor market will take; I’d guestimate it to be by a factor 2-4x, although to be fair I have no hard data for that. (For the sci-fi fans, it would reduce the transition time just as the Foundation was going to reduce the length of the dark ages before the Second Empire.) It will still be disruptive and scary for some, but the downward economic impact of the transition will be smaller and the resulting economic boom that follows technological changes will arrive sooner.
Articles about AI overlords may be great clickbait, but they distract us from the more likely and more immediate impact on the labor market. While long term AI is likely to cause more access and prosperity in the long run, as technology often does, the level of short-term disruption is unprecedented. Governments and society can minimize this impact if politicians can work to take the necessary steps, if they are willing to act. OK, now you can despair.
In part two we’ll examine the impact Ai will have on information and society.
It’s critical to learn about corporate culture before you accept a job offer but it can be awkward to raise such questions. Learn what to ask and how to ask it to avoid landing yourself in a bad situation.
Investing just a few hours per year will help you focus and advance in your career.
Groups with a high barrier to entry and high trust are often the most valuable groups to join.
