Artificial Intelligence, Part II – State of the Union

First and foremost, let me start by saying that AI will augment, not replace human labour. This is important. AI has the potential to dramatically improve our lifestyles, health, wealth and overall wellbeing. If managed appropriately, from the policy perspective, AI will create unlimited opportunities for us to live more fulfilling lives. Its transformational power, however, can go both ways. Our current environment, be it political or economic, is not suited to dealing with existential problems on a global scale. These problems require multilateral solutions. Climate change is one example. Our global response to coronavirus is another. In the environment where key parties are either unable or uninterested in bridging their ideological differences, the status quo is the only path of least resistance. So let’s recap the state of the nation, before the pandemic.

The employment picture in the US looks quite bright on the surface but the underlying trends are less encouraging. The participation rate has declined, leaving more and more Americans behind. Underemployment rate remains above 7%, indicating that many are discouraged, stopped looking for work or are working part-time but would prefer to work full-time. There is a gap between the skills employers are looking for and those possessed by the market. There is also a geographical mismatch between the jobs and the workers. Obama administration defined the impact of technological innovation on the job market as skill-biased technical change, favouring higher-skilled workers. As conveniently demonstrated by Amazon, selecting New York and suburban Washington, D.C. for its second headquarters, the necessary skills in today’s job market reside predominantly in metropolitan areas. Amazon pulling out of New York is a one-off, not a change in trend. The trend is exaggerated by slowing new business formation outside of metropolitan areas. Which consequently leads to an even greater rural vs. urban divide, as new and small businesses have historically created 68% of net jobs in the US [1]. To add insult to injury, antitrust enforcement is largely non-existent in the US. For example, 90% of America’s beer production is controlled by Molson Coors and AB InBev. Just four airlines control 70% of the market, with each having several regional hubs with limited competition. Same story in railroads, pharmacies, prescription drugs, search, digital advertising, social media, financial services, etc. Over 75% of U.S. industries have experienced an increase in concentration levels over the last two decades [2]. Furthermore, a number of publicly listed companies in the US is now less than 30 years ago, while the GDP has nearly tripled [3].

These underlying dynamics have led to higher margins, greater profitability and increased shareholder returns, while at the same time resulting in higher prices for consumers, lower wages for workers and lower R&D spending, all in favour of executive compensation and buybacks. Wages in the US have not gone up meaningfully in decades and inequality is at the levels last seen during the Great Depression. According to research from Gabriel Zucman, an economics professor at the University of California, Berkeley, the top 1% now owns about 40% of total household wealth. Further, the top 1% richest U.S. families own 40 times the average family’s wealth.

Thinking about AI-driven automation, we should consider which jobs are easiest to automate and what that might mean for society. According to the Bureau of Labor Statistics, retail salespeople, cashiers, fast food workers and truckers account for almost 12 million jobs out of a total 145 million. A large portion of these workers live below the poverty line and are recipients of government aid. Office clerks, waiters and customer service representatives account for another 7.3 million. These are the official numbers. The Aspen Institute, for example, estimated that nearly 16 million people worked in retail alone in 2017 [4]. They also found that over 80 per cent of workers in the retail industry do not have a college degree, and over half lack any post-secondary education. With that in mind, it is clear that AI can soon affect the most vulnerable population of American workers.

Obama administration, in its 2016 “Artificial Intelligence, Automation, and the Economy” report, placed a strong emphasis on education and training, including continued education, apprenticeships and re-training efforts. Indeed, millions of Americans will have to undergo additional training to prepare for the age of AI. Yet, about 44 million graduates already hold student debt of $1.5 trillion through the second quarter of 2018. According to Bloomberg, student loans have seen an almost 157 per cent cumulative increase over the last 11 years, while auto loans have grown 52 per cent. At the same time, the US government debt exceeded $24 trillion and is poised to grow further due to unprecedented fiscal stimulus to fight the coronavirus. It’s challenging to see where the funding for these programs to retrain millions of Americans in the next decade is going to come from.

The last thing I would like to bring up in describing the current environment is the nature of American politics. The revolving door between companies and regulators who are supposed to regulate those same companies is, frankly, ridiculous. Trump attempted to sign new ethics rules early in his presidency before being lobbied by his staff to drop the idea. Few Americans hear about this but the revolving door has been on the legislative agenda for a while with proposals covering increasing the lobbying ban, expanding the definition of lobbying, and forcing financial regulators to recuse themselves for two years on sensitive actions. Lobbying, on the other hand, has been extensively covered by the media yet no political action is on the horizon. Before lobbying became a lucrative business, less than 5 per cent of retiring members of Congress became a lobbyist. A 2016 study by three political scientists showed that 50% of retiring senators and a third of retiring House members now register as a lobbyist [5].

Adding to the concentration of political power in the hands of corporations and the wealthy is the emergence of super PACs and other dark money groups in the American political landscape. Keeping political donations anonymous, these groups don’t have to report their wealthy individual and corporate donors to the public, or the Federal Election Commission for that matter. In the 2016 election cycle, outside groups spent more than $1.4 billion, compared to $1 billion in 2012 and a mere $338 million spent in 2008.

This narrative about the state of the US economy, democracy and society is, of course, an opinion. Numbers and statistics are, and always will be, subject to interpretation. Judging by corporate profits, or the stock market or even the GDP, all is well in the USA. But looking at the wellbeing of the nation, it’s clear that the US has been going backwards for the better part of the 21st century. In the world of AI, these fundamental flaws will be exposed.

Part III will discuss why this time is different and why Industry 4.0 (or the fourth industrial revolution) is unlike any of the previous three industrial revolutions we have seen to date.

[1] St. Louis Fed, “Are Small Businesses the Biggest Producers of Jobs?”, Kevin L. Kliesen , Julia S. Maues
https://www.stlouisfed.org/publications/regional-economist/april-2011/are-small-businesses-the-biggest-producers-of-jobs

[2] “Are U.S. Industries Becoming More Concentrated?”, Gustavo Grullon, Yelena Larkin, and Roni Michaely
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2612047

[3] https://www.cnbc.com/2017/10/25/where-have-all-the-public-companies-gone.html

[4] https://www.aspeninstitute.org/blog-posts/industry-at-a-glance-the-future-of-retail/

[5] https://link.springer.com/article/10.1057%2Figa.2015.16