Automation, education, and the looming skills deficit

The present and future of American manufacturing (at Tesla).

For the past several years, automation has been top of mind for many. In academia, deep learning has driven rapid advances in artificial intelligence in consumer products, including personal digital assistants, semi-autonomous vehicles, and numerous other information industries like journalism and marketing. The press now writes regularly about the role of automation in eliminating jobs. There are clear limits to advances in AI, but as we’re all seeing daily, even incremental changes in the ability of algorithms to automate traditionally human labor can easily disrupt markets with the promise of productivity gains.

Fears of economic change through automation are not new. Back in 1956, the U.S. Senate released a congressional report on the possible effects of automation (broadly construed), predicting both positive and negative effects:

While the degree of automation made possible by modern science may well surpass the limits of present imagination, it is important to note that not all workers, indeed, only a relatively small, although conspicuous, fraction of the total labor force will be directly involved… However much we may welcome the fruits of advancing technology-however optimistic one may be that the problems of adjustment will not be serious-no one dare overlook or deny the fact that many individuals will suffer personal, mental, and physical hardships as the adjustments go forward.

Work has definitely changed in the past sixty years, resulting in many personal, mental, and physical hardships that emerged. But change occurred over decades, giving people who lost their jobs time to anticipate market changes, giving the government and employers time to subsidize retraining, and giving markets time to adapt to new technologies.

Now having similar debates about the coming tide of automation. The White House’s Office of Science and Technology Policy (under the Obama administration) wrote about the similar kinds of concerns last year:

One important concern arising from prior waves of automation, however, is the potential impact on certain types of jobs and sectors, and the resulting impacts on income inequality. Because AI has the potential to eliminate or drive down wages of some jobs, especially low- and medium-skill jobs, policy interventions will likely be needed to ensure that AI’s economic benefits are broadly shared and that inequality is diminished and not worsened as a consequence… Analysis by the White House Council of Economic Advisors (CEA) suggests that the negative effect of automation will be greatest on lower-wage jobs, and that there is a risk that AI-driven automation will increase the wage gap between less-educated and more educated workers, potentially increasing economic inequality.

Unlike the past sixty years, many have seen the change driven by computation move an order of magnitude faster. The U.S. Postal Service rapidly shrank after email was widely adopted. In just 15 years, all but the most widely read newspapers have declined, with many closing, resulting in dramatic job losses. Brick and mortar stores are being abandoned, displaced by online shopping, mostly at Amazon. Uber was just founded less than 8 years ago, and taxi companies gasping for air, while drivers in most cities are making less than they used to. And the coming automation in the next decade threatens to displace tens of millions of truck drivers, grocery store checkers, and fast food wage workers. And all of this will likely come in the next decade or two, with more changes to come that we haven’t even dreamt of yet.

These jobs will likely be replaced by a smaller number of higher paying, higher-skill jobs. That’s where education comes in: if we believe this predicted true, there are a variety of ways to react politically:

  • Allow the income inequality to occur. In the past, this level of inequality has led to civil unrest and sometimes civil war. I can’t see progressives tolerating this.
  • Mitigate inequality through social programs (including the more extreme ideas such as universal basic income). I can’t see conservatives tolerating this.
  • Low-income, low-skill citizens will have to learn new skills, competing for a diminishing pool of higher wage jobs. Given the United State’s embrace of capitalism, this seems like the most likely outcome.

Now, I’m typically an AI skeptic, but it doesn’t take fancy AI to disrupt markets. Amazon is a bunch of text, images, and web forms, plus some logistics automation. Uber is an app, some matchmaking algorithms, some demand-based pricing models, and a whole bunch of human resource and reputation systems. These are not bleeding-edge-deep-learning-machine-sentient innovations. These are simple software systems, built from decades-old ideas in computer science, economics, and HCI. If the truly bleeding-edge advances in machine intelligence achieve even a fraction of the hype in the press, we’re in for a long multi-decade, perhaps century long period of disruption.


The Whole Foods lines (and jobs) Amazon hopes to eliminate

What does all of this have to do with education? Take my predictions above as given for a moment, and imagine you work a low-wage job as a checker at a Whole Foods. The benefits are pretty good, the work isn’t too rough, and you like your co-workers. Amazon just bought your employer, and will inevitably try to increase productivity through automation—that’s Amazon’s secret sauce in all of it’s businesses, after all. You know your job will be going away someday, if not soon. What can you do if you want to find some stability for yourself and your family?

Let’s assume for a moment that you have savings, the confidence to learn new skills, and opportunities in your city to learn and secure a job in a new field (this describes only a tiny fraction of low-wage workers in America, but let’s tackle the ideal situation first). Even if you had these things, imagine the risks that learning new skills requires:

  • You have to have no income for a period of time while you learn
  • You have to successfully learn, which requires great teachers
  • You have to time your learning with the market demand for those skills
  • Ideally, you choose skills that pay at least as much as you’re paid now
  • You have to choose skills that aren’t likely to be automated soon
  • Most of all, you have to change your mindset from education being something you do while you’re young to something you do throughout your life.

Ultimately, the question moves from “What should I major in?” to “What should I major in next?

These are not simple problems. And the predictions above about the market aren’t something even economists can make accurately, let alone an individual. And if you’re someone with little education to begin with, many of these critical questions about your economic future may not even come to mind, as they’re ideas one might only be exposed to in college.

Now imagine the more typical case: you don’t have savings, you don’t have the confidence to learn new skills, and you don’t have opportunities to either learn or get new jobs in your city. What are you going to do?

This is the the rapidly approaching skill deficit. I predict that most low-wage Americans (i.e., most Americans) will have their jobs rapidly displaced by automation, and they will be trapped with no feasible way to learn marketable skills. And for those fortunate few who find a way, just after they acquire new skills, those will either be automated too, or the smaller labor market for those higher paying jobs will be quickly saturated.

In fact, this has basically already happened in much of rural America, especially in manufacturing. Manufacturing in America has resurged, but most of the jobs were lost to automation. The jobs that exist now require advanced training on configuring (i.e., programming and maintaining) robots.

What will this skill deficit mean for America? We’re beginning to see it’s effects already: a demagogue as president, rising nationalism, rising income inequality, and a raft of irrational, unsubstantiated theories about immigrants and religion as the root of our economic woes. Few suffering the effects of automation seem realize that what they really need is new skills, and more importantly, opportunities to learn new skills both rapidly and across a lifetime. Building walls and keeping immigrants out isn’t going to make anyone’s labor more marketable or relevant to our advanced economy.


2010 rallies in against cuts to public education in Minnesota (Credit: Fibbonaci Blue)

I know I’ve already been a real downer, but there’s more. At the exact same time that automation is necessitating lifelong learning, American politics are rapidly disinvesting in education at all levels. Dozens of states (including my own Washington State) have defunded public K-12 education to a point that state Supreme Courts are finding state legislatures in violation of their state constitutions. State investment in public universities has been in decline for decades. And now Trump is proposing a 40% cut on job training programs for adult learners. The clear trend is that politicians—and by implication the voting American public—wants less investment in learning at all ages.

This means that while Americans will be under further economic stress to continually pivot, retrain, and leap to new careers that have yet to be displaced by automation, there will be even less help and opportunity to do this. Moreover, because of the defunding of K-12 education, students at all but the best public schools will probably be even less prepared for college, less confident at learning, and less well-positioned economically to pursue the shrinking pool of middle-class jobs. America will need many more people with advanced skills, but the population with the desire to get these skills (including immigrants) will be on the decline.

Perhaps the most worrying trend is that even if funding for education was flat, that level of investment isn’t enough to keep great teachers engaged at their current salary levels. The current average teacher salary in the U.S. is $56,000, and the average starting salary is $30,000. The hours are long, the pay is terrible, the public gives you no respect, you start your day at 6 am, and unless you work in a wealthy district, the buildings are old, stuffy, and possibly crumbling. It’s no wonder that 50% of new teachers leave the profession within 5 years of starting. All of this means that enrollment in teacher training programs in the U.S. is down more than 50% in many states since 2010, threatening a teacher labor shortage, forcing school principals to either:

  1. Pay higher salaries to recruit from a smaller pool of teachers.
  2. Build larger classrooms to raise class sizes.
  3. Accept even less qualified teachers for declining pay.

Any predictions which it will be?


Is there any reason to be optimistic about this mess?

In the long view, I think so. Researchers like myself are working hard on discovering ways to make learning faster, more effective, and more scalable (ironically, through automation), which should help good teachers be great and great teachers be greater. We’ll have better tools for formatively assessing learning outcomes, which will both improve learning, but also make students struggling more visible to teachers, parents, and the public.

Outside of research, I’m optimistic that the public will eventually see that if we don’t commit to investing in people and their abilities, we’re really left with nothing: no consumers for businesses, no skills for new businesses, few skills for solving the inevitably harder problems that we’ll face in the world. We’ll have a small number of highly skilled people wanting to sell to a large number of people with no money to buy. At some point, people will remember that human ability is key to human prosperity, and we need to once again place learning at the heart of public investment.


But this future won’t happen automatically. People like you and and me who are passionate about learning, concerned about automation, and speaking this truth to power must stand up and demand this investment in learning. It’s the only way we all survive the dramatic economic change threatening to devour our country over the coming decades.

What are some concrete things you can do, especially if you have spare time or money? Advocate for publicly-subsidized education for all ages:

  • Vote to support public education and retraining on every ballot.
  • Vote for candidates that support public education and retraining.
  • Educate people in your networks about the importance of skills.
  • Make sure people in your networks actually vote.
  • Work in public education, as an educator, administrator, IT specialist, even if it means making less money. Become a CS teacher!
  • Give money to public education.
  • Volunteer at under-resourced schools (no, not the upper-middle class school your children attend).
  • Mentor people—children and adults—who need encouragement, resources, and support in learning new skills.
  • Advocate for research funding on education research of all kinds (especially computing education!), so we can make learning of advanced, hard-to-automate skills more effective, equitable, and scalable.

Any one of things may seem small, but if we all contribute a small amount, we can all make a huge difference in re-establishing human learning—and not machine learning—as the fundamental way that we fuel growth and prosperity in America and around the world.


P.S. While I’m blaming computation and machine learning for a lot in this piece, I don’t hate it. Code is handy for a wide range of problems in the world. It’s just not the solution to all of the world’s problems, despite what the press and computer scientists say. Code a very specialized tool that hammers very specialized nails, and only when handled with extreme care and precision. Let’s use it where appropriate and, where not, leave decisions to humanity.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.