The Startup
Published in

The Startup

It’s Time to Face Automation

Automation is a growing threat to job stability around the globe.

It is widely recognized that the rise of artificial intelligence and automation will make the production of products and the supplying of services more efficient, as computers start doing what previously took hours in minutes. Already, we can trace ~15% of recent productivity growth to the use of industrial robots. (Graetz & Michaels, 2018, p. 754) However, it is increasingly clear that the integration of these machines will result in lost jobs, as people who are getting replaced cannot get new jobs in similar fields. (Das & Hilgenstock, 2018, pp. 33–34) The confluence of these opposing forces poses huge challenges to the stability of the middle class in particular and modern capitalism as a whole. So what if we do nothing about it, and how can we stop it from happening?

As humanity progresses, the amount of people that need to be doing raw labor decreases; ten million people worked in agriculture in 1950, compared to just three million that work on farms today. (United States Department of Agriculture National Agricultural Statistics Service, 2017) Conversely, the number of people doing more complex forms of labor has increased, such as in the fields of finance, transportation, and administration. Thus, the percentage of people employed has stayed relatively stable. This is the automation of physical work, and has not caused civilization-breaking problems yet. What we’re seeing now, however, is the automation of intellectual work. How exactly do you do that?

There are a multitude of ways to create useful AI. The easiest to explain involves creating lots of robots and then breeding, testing, and culling them. This is called “genetic breeding,” because of its similarities to natural selection: the robots have random mutations, and the ones with the most useful random mutations survive. Other methods have names like Q-learning, which utilizes rewards in a virtual space, and deep learning, which is akin to operating the world’s most confusing slot machine. (Sutton & Barto, 2018) The end result of all of these methods is automation of intellectual work. That’s how most technology that would seem like magic 30 years ago, such as Google’s Search and Tesla’s Autopilot, works. This type of automation reduces the amount of intellectual labor required without generating significant numbers of new jobs. Apple provides 2,000,000 jobs and is worth over $1 trillion, meaning a value of $500,000 per employee. (Apple, 2016) Ford employs 190,000 people and is only worth $35 billion, resulting in a value of about $200,000 per employee. (Forbes, 2020) In the age of AI, you need less people than ever before to generate more money than ever before- productivity is becoming decoupled from labor.

What Will Happen

Lots of ink has been spilled in arguing just how much AI and automation as a whole will impact the global labor market. As Acemoglu and Restrepo wrote (2018),

On the one side are the alarmist arguments that the oncoming advances in AI and robotics will spell the end of work by humans, while many economists on the other side claim that because technological breakthroughs in the past have eventually increased the demand for labor and wages, there is no reason to be concerned that this time will be any different. (p. 1)

Winick (2018) compiled studies estimating job losses in the upcoming 25 years due to AI. The conclusion drawn was that while most agreed jobs would be lost, no studies agreed on even an approximate amount. Estimates for job losses ranged from about 2,000,000 on the low end (by 2020) to 400,000,000 on the high end (by 2030).

One thing is clear, however: companies and universities are investing heavily in Artificial Intelligence research. The non-profit organization AI Index (2017) found that the number of papers written about AI ballooned to 18,000 in 2017. Furthermore, the McKinsey Global Institute (2018) estimated that companies such as Facebook and Google have spent $18 — $27 billion dollars internally in the field.

We’ve established that the widespread use of AI is coming, and that when it does, many jobs will no longer exist. Now, we must determine what impacts this will have on the economy.

  1. Lower-income jobs are more likely to be automated before higher-salaried positions. This intuitively makes sense — a nascent AI will get good at filling prescriptions before getting good at writing them. Thus, as Prettner and Strulik argued (2017), we can predict a rise in inequality. (p. 250)
  2. Because AI is only as good as the data it’s built on, we may see lowered competition as companies with giant datasets are able to make the best AI models. (Goolsbee, 2018, p. 5) Historically, small businesses contribute greatly to productivity growth, as increased competition, especially on small scales, drives innovation. AI directly stifles that, by erecting a high barrier to entry in the form of a high-quality data set.
  3. Privacy will be eroded. We already are seeing this today, with massive and controversial collection of data by companies like Facebook, Google, and Twitter.

You may notice that modern capitalism relies on many of the things AI is expected to have an adverse effect on. Capitalism works best with a large middle class, and AI will increase inequality. Capitalism depends on abundant competition, and AI is most efficient when used at large scales. Capitalism only works when consumers have a choice, but if advertisers can build nearly perfect models of your psychology, can you really be certain you chose to buy that box of cereal? Because of this, automation may reduce the usefulness of capitalism as a vehicle to raise standards of living both in reality and in the eyes of the public.

What to do

Overall, if nothing is done to blunt the most serious negative impacts of AI, we can expect major job losses, rising inequality, and eroded privacy. What can we do to dull this? The most urgent problem is the impending hemorrhaging of jobs. One solution that promises to prop up the economy is Universal Basic Income (UBI). The basic idea is simple enough: replace most welfare with a government-provided check that goes to all citizens every month. While at first this proposal seems simple and logical, upon closer examination, problems become apparent. For now, we’ll ignore the political challenges and consequences of axing most welfare. The most obvious drawback is the immense cost UBI would place on taxpayers. Hoynes and Rothstein (2019) show that a UBI that provides basic support for a household would cost “about twice the cost of all existing [welfare] transfers in the United States.” (p. 954) The question of how these funds are raised will play a large part in determining the effectiveness of UBI. In addition, because UBI is not means-tested, it will be more regressive when compared with the current social safety net. It is reasonable to infer, then, that UBI may simply add to the inequality intellectual automation will cause.

Instead of softening the impact of automation, Bill Gates and others have proposed a “robot tax,” which would penalize automation directly. Gasteiger and Prettner (2017) have shown that a robot tax on its own would not eliminate stagnation in the labor market on its own, but could increase per-capita capital. Furthermore, they raise the issue of tax avoidance. It is likely that a robot tax would only be effective if a large number of countries simultaneously implemented it. Otherwise, an enterprising company can simply move the location of its programs and machines, either on paper or in actuality, to a country where they won’t be taxed at all.

Another large challenge when confronting automation will be the immense need for worker retraining. When workers are inevitably displaced due to both intellectual and industrial automation, they will have to gain new skills in order to stay in the labor market. Failing to provide adequate education facilities for these citizens will lead to more unnecessary unemployment. Jiamovich et al. (2020) used a machine-generated model to predict the impacts of widespread retraining programs. They found that in order to reverse “low-skilled” employment back to 1989 pre-automation levels, we would need to successfully reincorporate 10% of the unemployed. Obviously, this is no easy task. Multiple approaches have been suggested, from government-funded education, to corporate training programs. Even so, as Brynjolfsson and McAfee (2011) explain, due to the nature of exponential growth, it is probable that training will get harder over time, as AI is able to automate more and more jobs. In other words, we need to start right now.

In conclusion, automation forces us to confront many simultaneous serious situations. At the same time, it promises huge gains in efficiency, standards of living, and productivity. We have the ability to blunt the worst of the negative impacts, while still taking advantage of the positive ones, by using a variety of tools. And yet, right now, we aren’t heavily investing in any of them. For the first time in human history, productivity is becoming separate from labor. It’s time we decide, as a society, how to deal with that.


Acemoglu, D., & Restrepo, P. (2018). Artificial Intelligence, Automation and Work. National Bureau of Economic Research Working Paper Series.

Artificial Intelligence Index. (2017). 2017 Annual Report.

Autor, D. H., Katz, L. F., & Kearney, M. S. (2006). The Polarization of the U.S. Labor Market. National Bureau of Economic Research Working Paper Series, №11986.

Brynjolfsson, E., & McAfee, A. (2011). Race against the machine: How the digital revolution is accelerating innovation, driving productivity, and irreversibly transforming employment and the economy. Brynjolfsson and McAfee.

Bughin, J., Seong, J., Manyika, J., Chui, M., & Joshi, R. (2018). Notes from the AI Frontier: Modeling the Impact of AI on the World Economy. McKinsey Global Institute.

Campa, R. (2017). Automation, Education, Unemployment: A Scenario Analysis. Studia Paedagogica Ignatiana, 20(1), 23–39. doi:

Das, M., & Hilgenstock, B. (2018). The Exposure to Routinization: Labor Market Implications for Developed and Developing Economies. IMF Working Papers, 18.

Executive Office of the President. (2016). Artificial Intelligence, Automation, and the Economy.

Ford Motor (F). (2020). In Forbes. Forbes.

Furman, J., & Seamans, R. (2018). AI and the Economy. National Bureau of Economic Research Working Paper Series.

Gasteiger, E., & Prettner, K. (2017). A Note on Automation, Stagnation, and the Implications of a Robot Tax. Diskussionsbeiträge Des Fachbereichs Wirtschaftswissenschaft Der Freien Universität Berlin, 17.

Goolsbee, A. (2018). Public Policy in an AI Economy. National Bureau of Economic Research Working Paper Series.

Graetz, G., & Michaels, G. (2018). Robots at Work. The Review of Economics and Statistics, 100(5), 753–768.

Hoynes, H., & Rothstein, J. (2019). Universal Basic Income in the United States and Advanced Countries. Annu. Rev. Econ., 11(1), 929–958.

Jaimovich, N., Saporta-Eksten, I., Siu, H. E., & Yedid-Levi, Y. (2020). The Macroeconomics of Automation: Data, Theory, and Policy Analysis. National Bureau of Economic Research Working Paper Series, №27122.

Job Creation — Apple. (2016). In Apple.

McKinsey Global Institute. (2013). Disruptive technologies: Advances that will transform life, business, and the global economy.

Pham, Q. C., Madhavan, R., Righetti, L., Smart, W., & Chatila, R. (2018). The Impact of Robotics and Automation on Working Conditions and Employment [Ethical, Legal, and Societal Issues]. IEEE Robotics & Automation Magazine, 25(2), 126–128.

Prettner, K., & Strulik, H. (2019). Innovation, automation, and inequality: Policy challenges in the race against the machine. Journal of Monetary Economics, 116, 249–265.

Prettner, K., & Strulik, H. (2017). The lost race against the machine: Automation, education and inequality in an R&D-based growth model PDF Logo. Hohenheim Discussion Papers in Business, Economics and Social Sciences, №08–2017.

Sutton, R. S., & Barto, A. G. (2018). Reinforcement Learning. Bradford Books.

The Diamond OLG Model. (2020). In Krieger School of Arts and Sciences. Krieger School of Arts and Sciences | Johns Hopkins University.

USDA/NASS Census of Agriculture Chapter 1, Table 1. (2017). In USDA — National Agricultural Statistics Service. USDA.

Winick, E. (2018, January 25). Every study we could find on what automation will do to jobs, in one chart.




Get smarter at building your thing. Follow to join The Startup’s +8 million monthly readers & +756K followers.

Recommended from Medium

Building a chatbot using Dialogflow

Is more Artificial intelligence good or bad?

CHI’21 Preprint Collection HCI + AI

AI Impact Control Panel

Where Humans Meet Machines: Intuition, Expertise and Learning

Trustworthy AI: Imagining Better Machine Decision Making

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Parv M.

Parv M.

A student who does fun things every once in a while.

More from Medium

The Attention Wave

What is the New Payments Architecture and what impact will it have?

$300,000 Is Not Paycheck To Paycheck In Silicon Valley

22 Lessons From 22 Months Building A Ghost Town