Betting against techno-unemployment

There are millions of truck drivers in the US today. In particular, there are about 1.7 million tractor-trailer (human) drivers. There are many more professional truck drivers in the US, but tractor-trailer drivers are the “archetypal truck drivers”.

In 2016, we had several demonstrations of self-driving tractor-trailer trucks. A self-driving tractor-trailer truck delivered beer for Uber in October.

Putting two and two together, many people are worried that we are soon going to face a wave of technology-related unemployment. You can’t, realistically, retrain over a million tractor-trailer drivers into Facebook engineers.

However, when jobs disappear slowly, it does not always translate into unemployment. For example, people tend to leave naturally an industry even when they still have a job. This occurs through retirements and deaths, but also in various other ways. Maybe someone gets tired of the long hours involved when driving trucks, and they go for an office job. Some people decide to start a small business. And so forth. Moreover, when the job prospects are poor, fewer people tend to enter an industry. There may be no need to fire anyone.

This kind of attrition is likely happening to journalism today. In the past, every town had its independent newspaper. Newspapers were our primary information source about the world. Today, fewer and fewer of us read newspapers and those who do tend to be older. The bulk of the journalism jobs were never at the New York Times, they were always in a myriad of small papers. These small papers are mostly obsolete. Many prospered through classified ads, but newspapers classified ads are a relic of the past. The few small newspapers that remain can be run with a much smaller crew. So how did the workforce fare? We went from half a million (in 1990) to less than half that (180,000) in 2016. Of course, over the same period of time, about 200,000 new jobs were created in Internet publishing, a closely related industry. So while there are fewer journalists today, it took 16 years to see the disappearance of 2/3 of all jobs… and the Internet created many related new jobs. All in all, it was not a good time to be a journalist, but it did not create a national emergency.

Journalism has been something of an extreme case. The Internet really disrupted journalism. If we look at other industries, like the automotive industry, we see much more modest job losses (like 20% over 15 years). Such modest job losses are still enough to create some pain, but most employees did fine.

There are many more truck drivers than there ever were journalists. It is almost an order of magnitude in difference. And while journalists can find new jobs in the vast fields of communications and entertainment… I don’t think anyone would know what to make of a million unemployed truck drivers.

Yet, even if, eventually, all tractor-trailer drivers are computers, it does not follow that we will experience techno-unemployment in the sense that those in the profession will see their careers destroyed. We need a sudden fall in the number of jobs.

Calum Chace wrote an excellent book “The Economic Singularity: Artificial intelligence and the death of capitalism“. As the title of his book suggests, Calum predicts some major disruption.

In particular, Calum predicted the following in an email to me:

By 2030, there will be fewer than 250,000 tractor-trailing drivers in the US, according to official government statistics.

The year 2030 is 13 years away… To fall down to 250,000 drivers from 1.7 million in 13 years, we would need to experience a 13% decline every year until then… It is the equivalent of the disappearance of 85% of all jobs… something much more severe than what journalism underwent. Something much, much more severe than what the automotive industry underwent.

I have decided to place a bet against Calum. He wins if during any year, up till the year 2030 (inclusively), the number of drivers fall under 250,000. I win otherwise. We each put $100 on the bet, with the money going to a charity chosen by the winner.

Though a nice theory, techno-unemployment is not currently occuring, despite all the technological progress we see. It is an idea, but not necessarily something that we can observe in the real world.

I think that there are a few good reasons for why that is.

  • There is a wide gap between what is technically possible and what we can achieve in the field. For example, we could pretty much automate at a large scale how medical data in handled. Yet very few countries did it, and few countries are in the process of doing it. Let us recall the healthcare.org fiasco where the US federal government deployed a fairly simple web application to help people find affordable insurance.

    So even if we can build and sell perfectly safe self-driving tractor-trailer trucks today… one should not underestimate the difficulties that we might have to deploy this technology. In fact, I can imagine a scenario where 2030 come and pass, we still do not see self-driving tractor-trailer trucks on the roads as a matter of course.

    Did you know that New Jersey and Oregon, two otherwise reasonable American states, ban self-service gas stations? Yet self-service gas stations were aggressively introduced back in the 1970s. How long do you think it will be before all American states allow self-driving tractor-trailer trucks?

    As far as I can tell, subways worldwide could easily be automated. It is simply not that hard to automate a subway train. Yet, according to Wikipedia, there are only two automated train networks in Canada.

  • Even when good automation technology does get deployed… if the industry grows sufficiently fast, it leaves ample room for professionals to earn a good living. For example, there are more human bank tellers employed in the US right now than there ever were. Part of it is due to regulations, but it would be impossible without a massive expansion of the banking industry.

    It is entirely possible that automation will be used mostly to grow the industry. There could still be a million truck drivers by 2030, but maybe they will focus on the less frequent routes while the most intense routes are automated.

Generally speaking, my view is that we are not going to be able to automate fast enough. In most high-income countries, we have aging populations. We are going to run out of nurses willing to work for wages people can afford. I am hoping that health care will improve at an accelerated rate so that we can start getting the diseases of aging under control in 20 years… but even if we do, we simply won’t have an ample supply of new eager employees.

My thesis is that techno-unemployment should be the last of our worries. Soon enough, we will live in countries where a quarter of the population is at an age where we expect people to be retired. It is already the case in Japan.

On the one hand, this should please people who fear overpopulation, as it means decreasing population (people over the age of 65 tends to have few children). On the other hand, this means few young new employees. We will need all the robots we can get.

We need self-driving trucks because we are going to run out of people to drive them.

12 thoughts on “Betting against techno-unemployment”

    1. @Dominic

      I think that in a lot of industries, that’s exactly what is likely to happen, but it is not mere coincidence.

      For example, the US must import nurses from the Philipines right now. It is like we are going to have an overabundance of nurses any time soon.

      So we can’t wait to automate some of what nurses do. It is unlikely that we’ll have fully automated nurses by 2030… but we probably can’t get some more automation fast enough.

  1. Automating subways is surprisingly hard. Helsinki has already tried it twice without success. The first time was in the 70s, when the technology was simply not mature enough. The second attempt started in 2006 and was abandoned in 2015, when it had become obvious that the company that won the contract could not make the existing trains run automatically.

    1. I don’t think it is hard to automate subway trains in 2017 for technological reasons. We have had fully automated subway trains in some major locations for many years now… Actually making it happen is obviously much harder. And that’s my point regarding automated trucks. Even if we can automate them today, without strong incentives, it is going to be hard to deploy them at scale.

      Regarding subway trains, I can see many challenges. For one thing, the financial incentive is not huge. You save one person’s salary… so maybe 100k$ a year. But you still need supervisors, mechanicians and so forth. To top it off, you need extra capital investment and probably extra staff to manage the new computers. It is very hard to see how you can make a lot of money by automating subway trains unless you do it at a large scale… and at that point, you have other problems.

      1. Building a new automatic subway system is certainly doable. Automating an existing one is much harder.

        A subway is essentially a massive one-of-a-kind legacy system. Automatization requires making major changes to the system while keeping it running all the time. The technology to do it exists, but the ability to do it reliably does not.

  2. Great article. Can I get on your side of the bet?

    Looking at current technologies that have existed for years/decades yet are still not being implemented across the board should show how hard it will be to implement driverless trucks.

    I’m no expert, but from what I know, there is little reason for a truck to not have a skirt under the trailer, as it pays off in 3-4 years and then provides significant fuel savings, yet I see many, many trucks without them.

    There are similar advances without quite as large a benefit, but with benefits nonetheless. The trailer tail and wide single tires instead of duals are still not even close to being adopted after being available for over a decade.

    Soon, driverless trucks will almost certainly look good on paper for a larger company flush with cash and willingness to take a risk, but that’s a lot different than large-scale implementation.

    1. Can I get on your side of the bet?

      Calum may be willing to entertain further bets.

      Looking at current technologies that have existed for years/decades yet are still not being implemented across the board should show how hard it will be to implement driverless trucks.

      It is certainly uneven. I think it comes down to incentives, politics and all sorts of soft considerations that have little to do with hard science. All these factors are hard to predict.

    2. I’d like to get in on Calum’s side of the bet, even if 2030 is actually only 13 years away.

      I’d say that, by 2030, at the very least all long-haul trucking will be automated, if only because of safety improvements and the ability to run the trucks 24 hours per day.

      That’s assuming they’re not replaced by Amazon blimp warehouses and drones. :^)

      1. I’d like to get in on Calum’s side of the bet, even if 2030 is actually only 13 years away.

        You’d think I could do basic arithmetic by now.

        I’d say that, by 2030, at the very least all long-haul trucking will be automated, if only because of safety improvements and the ability to run the trucks 24 hours per day.

        What are your thoughts on subways? Do you think that major cities in the US will have fully automated subways?

        That’s assuming they’re not replaced by Amazon blimp warehouses and drones.

        Even if all Amazon’s transportations needs are met with drones, it still won’t cut it.

  3. > [In] the automotive industry […] most employees did fine.

    So you say Detroit did fine.

    > I don’t think anyone would know what to make of a million unemployed truck drivers.

    That’s one problem. You give counterexamples of journalists, bank clerks, automotive workers and nurses, but they either are in better position because of education, are in growing business but not fit for men or didn’t work that well.

    Second problem is with all the countryside infrastructure which will collapse – just like in the Pixar’s Cars movie.

    > I can imagine a scenario where 2030 come and pass, we still do not see elf-driving tractor-trailer trucks on the roads

    Sure, automotive driving is far away, probably further than 2030, but we know it will happen – just like it happened with agriculture and automotive. And we should prepare for it, rather than neglecting and trivializing.

    1. So you say Detroit did fine.

      According to the US government, in 2009, blue-collar workers in the automobile industry received 37$/h in compensation. (Source) That’s between 70k$ to 80k$ a year. I know many college graduates who are not compensated that well. There are fewer workers today in the US, but only about 20% fewer… and the decline was over many decades.

      Regarding Detroit… Entire books have been written about the rise and decline of Detroit. It is a complicated story. The following passage from Wikipedia gives us a clue…

      These processes, in which the growth of the auto industry had played such a large part, combined with racial segregation to give Detroit, by 1960, its particularly noteworthy character of a substantially African-American inner city surrounded by mainly white outer sections of the city and suburbs. By 1960 there were more whites living in the city’s suburbs than the city itself. On the other hand, there were very few African-Americans in the suburbs. Real estate agents would not sell to them, and if African-Americans did try to move into suburbs there was “intense hostility and often violence” in reaction.

      To sum it up, the people who held the good jobs in the automotive industry did not live in Detroit proper. If you look today at the suburbs of Detroit… even today… I think you will find that they are doing ok.

      Sure, automotive driving is far away, probably further than 2030, but we know it will happen – just like it happened with agriculture and automotive. And we should prepare for it, rather than neglecting and trivializing.

      The purpose of this bet is to get people to think about the future.

  4. I think the real problem is that today there’s more innovation in educating computers than educating High school students in Europe and the USA. While there’s no real AI risk to humans there’s a really serious problem with the lack of innovation in High school education. In fact, the growing risk of severe long-term unemployment in these regions isn’t due to the arrival of super-human robots which won’t happen anytime soon, it’s due to a complacent education system that hasn’t taken into account significant changes in the modern job market.

    I wrote more about this here: https://keplerlounge.com/2017/01/16/where-education-failed/

Leave a Reply

Your email address will not be published. Required fields are marked *