Lessons Learned from a Summer Binge Reading About the Impact of Artificial Intelligence on Work and Society




Spend a couple of months immersed in books, news articles, and studies on how artificial intelligence is changing the workplace, and it’s hard not to conclude that a comprehensive rethinking of our societal institutions is in order.  Yes, these startling technologies are turning science fiction into reality.  Yes, they will improve our health, make our lives more secure, and vastly increase the efficiency of business, government and society. But there will be less sanguine changes as well.

The development of artificial intelligence is nearing an inflection point such that improvements will be accelerating at exponential rates, increasing productivity and potentially putting millions out of work.  Will that be a blessing or a curse, or both?  Will the ranks of the long-term unemployed further deepen, or will advanced automation come to the rescue as the labor participation rate continues to fall, boomers retire, and the job market tightens?  Will our tax, health, training, employment, income assistance, and other structures fit this new era?   All these questions create public policy issues with enormous implications for employers.

Regarding the future of work, the debate is not about whether AI will cause displacement.  Rather, it is about its extent, rapidity, and the solutions for its mitigation. Of even more importance, this displacement is not speculation by academics, screen writers, reporters and consultants.  It is what business leaders themselves are saying.

In a recent survey of HR Policy Association members, all of whom are large employers, 23 percent said that automation will replace up to 10 percent of their current workforce by 2030, 39 percent said up to 30 percent and 10 percent said as many as 50 percent of their company’s staff will no longer be needed.  

I asked our Chief Economist, Mark Wilson, what numbers like these really mean.  He said that if they are extrapolated across all large employers, between 4.3 and 11.1 million Americans may have their jobs eliminated by the end of the next decade.  He further suggested when keeping in mind historical trends in BLS data on average monthly job growth, it can be said that 13.3 million jobs will be created by 2030.  However, factoring in the AI job loss indicated by the survey and extrapolating that loss across all employers with more than 100 employees, we may see 47 to 118 percent of all new net jobs that otherwise would be created by 2030 simply vanish.  That means when the Labor Department does its monthly announcement of the unemployment rate and job figures in the next decade, there may be times when no new jobs are reported.  More detail regarding his projections can be found here.

Previous economic revolutions have demonstrated that while some workers lose their jobs, far more opportunities are created in new career fields.  We have seen that appropriate training and upskilling along with the passage of time help the displaced move into new occupations. With the AI revolution, however, one question is whether that historical trend will hold true.  Will there be enough new occupations to absorb those affected?  How much difficulty will the displaced have being repositioned?  Will some be pushed out of the job market altogether?  This same survey posited that as machines acquire greater capability to take on cognitively complex tasks, there will be significantly less need for humans in the workplace. Only 33 percent disagreed with the statement.   

We know that any new technology implemented on a broad scale tends to set off alarms at its introduction but is assimilated over time.  Still, from a public policy perspective, there are caveats with the coming AI revolution.  First, the world has never gone through a major industrial change during a time when social media, the internet, and 24/7 news were ubiquitous with most everyone walking around carrying a camera, video recorder, and microphone capable of recording and publishing content globally.  If there is pain, we will hear about it.  Second, the steam engine which powered the first industrial revolution didn’t come with machine learning and connectivity to the web enabling it to improve its capabilities autonomously and take actions not contemplated by its inventors.  Yes, automation has been a key element of all industrial revolutions, but a type of automation capable of perceiving, evaluating, reasoning, communicating, and engaging in self-transformation will be something altogether different.

For these reasons, it may be important to begin an exploration of the impact of artificial intelligence on work and determine which public policy issues may need to be addressed.  The following are some worthy of consideration.     

What does AI replace?

The point is frequently made that AI will not displace occupations as much as it will transform the processes used to get work done.  As described in a New York Times article, "Learning to Love our Robot Co-Workers," people will be realigned to work with new machines as they have since the beginning of the first industrial revolution to enhance their capabilities, achieve greater output, and create new job opportunities.  Dangerous jobs humans have been performing can be taken over by machines.  Still, a lot of people are going to have their jobs eliminated.  McKinsey has done several studies on this subject, finding that currently demonstrated technologies could automate 45 percent of all the activities people are paid to perform.  One study found that 59 percent of all manufacturing activities can be automated, 73 percent of activities in accommodations and food service, and 53 percent in retail.  Said another way, 60 percent of all occupations could see 30 percent or more of their constituent activities automated, according to McKinsey.

That’s why Erik Brynjolfsson and Andrew McAfee caution in The Second Machine Age, “There has never been a worse time to be a worker with only ‘ordinary’ skills and abilities to offer, because computers, robots, and other digital technologies are acquiring these skills and abilities at an extraordinary rate.”  A Washington Post story published earlier this month illustrates what happens when a low-tech manufacturing plant in a rural area struggles to staff its lines.  Giving up on recruiting, the owners begin installing robots alongside humans, the company finding the robots less expensive than even the lowest-paid worker plus capable of operating at a far faster pace.  The clear inference in the story is that the robots will be replacing additional production workers in that plant.  

But not just production workers are threatened.  The literature describes how automation will impact any job that is routine, predictable and can be reduced to a formulation.  Already, many large companies have instituted administrative and financial systems using robotic processes that have eliminated dozens of employees.  Machines are even penning financial and sports news as well as business reports.  Click here for an example that will give pause to anyone who writes for a living.

How fast will the changes come?

The U.S. may have been fortunate with previous economic revolutions.  For example, 70 to 80 percent of the workforce was employed in agriculture in 1870 while less than 2 percent was by 2008.  That nearly 140-year period helped mitigate cataclysmic dislocations.  However, with the exponential growth in AI applications, this change could come far more rapidly with little time to adjust.  According to Jerry Kaplan in Artificial Intelligence, “as long as these effects are gradual, the labor markets can adapt gracefully, but if they are rapid or abrupt, significant dislocations occur.”

Are changes necessary in our economic system?

The literature discusses how our economy is based primarily on people purchasing goods and services, with individual consumer spending constituting two-thirds of GDP in the U.S. and around 60 percent in most other developed countries.  In addition, major government programs are funded largely by payroll taxes.  If AI results in far fewer people working, then who will purchase what the machines produce and pay the taxes necessary for government to function?  And if companies are incentivized by tax, demographic and other factors to substitute machines for humans, will a labor-based tax system remain appropriate?  Some manufacturers and users of robots say things such as “we are engineering the labor out of the product” and “our device isn’t meant to make employees more efficient, it’s meant to completely obviate them.”  As more work is done by machines and less by humans, calls are being made to consider changing tax policy away from labor and more towards capital.  Bill Gates, for example, believes a company’s use of robots should be taxed to both slow the spread of automation and fund other forms of employment.  Both Tennessee and Massachusetts are already considering a per-mile tax on self-driving cars.

What regulation may be necessary?

“If there is one myth regarding computer technology that ought to be swept into the dustbin,” Martin Ford writes in Rise of the Robots, “it is the pervasive belief that computers can only do what they are specifically programmed to do.”  In fact, there are many, including Elon Musk and Stephen Hawking, who believe that AI represents an existential risk to the existence of human civilization.  They fear that machine learning will lead to singularity, the point at which AI produces runaway technological growth in the form of a superintelligence beyond the ability of humans to decipher or control.  There are others, such as Mark Zuckerberg, who are far less worried, but even Facebook is under fire as are other tech giants with threats of regulation as utilities.  The nanotechnology community is concerned that its ability to engineer at the atomic level may produce applications that could get totally out of control. The Chairwoman of the Federal Trade Commission recently announced its interest in artificial intelligence “because it has a consumer protection element to it but also has a competition element to it.” Therefore, it seems inevitable that regulation of these new technologies will begin in earnest and that will sweep in an examination of their implications on employment.  

How should displacement be addressed?

Among those assuming that AI will displace millions of Americans, there is a debate underway as to appropriate responses.  The idea most often discussed is creation of a Universal Basic Income, or UBI.  The payment could come in a variety of forms, such as a negative income tax, a different form of EITC, or a straight payment like Social Security.  Proposals are being made that everyone between the ages of 18 and 65 receive a government payment without any means testing or other evaluation regarding individual circumstances.  The idea is to provide a basic income floor that provides a minimum standard of living.  The hope is that people will use this floor to improve their circumstances, but most proponents accept the possibility that some portion of the population will choose to drop out and play video games all day.

The most eloquent proponent of the UBI is former SEIU president Andy Stern who has written Raising the Floor: How a Universal Basic Income Can Renew Our Economy and Rebuild the American Dream.  

The idea would be to use UBI to replace dozens of poverty programs now in place—welfare, food stamps, tax credits, unemployment comp, workers comp, minimum wage, housing assistance, and SSDI, for example.  However, that by itself would not create sufficient revenue to provide a minimum annual payment of $10,000, the number most often proposed.  Additional taxes would be required, and any policy that encourages people not to work and requires higher taxes promises to be highly controversial.  Also, nothing emerges from the legislative process without conditions being attached.  For example, do some individuals get a different level of benefit if they participate in certain types of training programs, engage in community service, or stay off drugs?  And what size army of bureaucrats would be required to police that?  Still, Stern has found people on both the right and the left who support the concept—“Conservatives see it as a means of eliminating costly and inefficient welfare programs.  Libertarians view it as a way to encourage greater individualism and personal choice.”  

What should employers be considering?

Billions are being spent developing the next generation of automation but virtually nothing on dealing with its consequences.  MIT, for example, has a cutting-edge research institute called the Computer Science and Artificial Intelligence Laboratory funded by a broad range of companies interested in supporting advances in AI technology.  Should a similar type of institute be developed as machines gradually consume the base of the skills pyramid, one that brings together a broad range of disciplines to study and come up with recommendations for policy initiatives?  As part of that, we seem to be at a point where the very concept of work itself is being redefined.  The authors of The Second Machine Age write that more than money, work is “one of the principal ways [people] get many other important things: self-worth, community, engagement, healthy values, structure, and dignity to name a few.”  What happens to our society when there are fewer avenues through work to achieve these important things?  Finally, we already have tremendous political, social and income divides in this country.  What will be the impact of AI?  

All these are important questions for employers to consider well before the political process is triggered.