This time is different
When Sir John Templeton coined the phrase – “The most dangerous words in investing are: ‘This time is different’” – he was not thinking about AI. In fact, Templeton uttered these fateful words in October 1987, just a short time before the Black Monday stock market crash. For nearly the past 40 years, Templeton’s words have been a warning to those who ignore history and precedent and instead assume things may play out differently this time than they have previously – often to one’s financial detriment.
The AI “doomers” have now adopted Templeton’s words, arguing that in fact this time is different. That is, the doomers claim AI will destroy jobs – and potentially destroy civilization – reducing us to a labor-less society in which we will have gobs of free time but without delivering significant enough economic growth to sustain our society.
Ok – I am exaggerating a bit for effect – but the arguments are not far off from this. But let’s focus on the core job destruction argument because that is the key doomer worry – AI will inevitably cause unemployment to spike and standards of living to fall precipitously.
I also understand from talking to many of my OPM colleagues that this is of concern to some of you. It’s completely understandable to feel angst about this topic – impending change often fosters such feelings. My goal with this post is to try offer a historical perspective on the jobs question. I don’t pretend to know all of the answers – and reasonable minds disagree on this topic – but hopefully this provides some food for thought and shows the opportunity we have to better serve our customers and the American people.
A quick disclaimer before we go further. I am trying to make a balanced historical argument in this post, but admittedly I am biased by my experiences. I have spent my entire professional career in the technology industry – as an operator and investor – and believe strongly in the power of technology, both to do well and (in the wrong hands) to facilitate evil. Technology is not the panacea for all of the world’s problems, but I do believe technological innovation is a key factor in our ability to enjoy long-term economic, job and productivity growth, stable economies, consumer surplus, and national security. So, you’ve been warned!
Back to jobs.
If the doomers believe this time is different, then what has been the actual historical course of job displacement during prior technology revolutions?
Economist Carlota Perez has developed what I think is the most interesting articulation of what she terms “technological revolutions:” sets of radical innovations that dramatically increase productivity. In each revolution, the pattern is fairly consistent.
In the first phase (she calls it “Installation”) new technologies arrive with great fanfare but also skepticism from the old guard. This phase typically drives speculative investment, culminating in a financial crash (think, 1929 or the 2000 Dot Com bust). Out of the ashes of the bust comes phase two (the “Deployment” phase), during which companies and consumers experience a golden age of economic and productivity growth. And in that golden age, we have consistently seen significant net job creation (and higher productivity, economic growth, and higher standards of living).
In the First Industrial Revolution, for example, some artisans who made clothing and crafts by hand were displaced, but a multitude of new jobs were created – factory operators, machinists, railroad engineers and conductors, etc. Recall this is why the original English “Luddites” – who were mostly textile artisans – smashed new machinery that was being utilized in various textile factories; they were understandably afraid of job displacement and hoped to stop it in its tracks by disabling the new, offending technology.
We have seen the same overall net job creation in each of the subsequent technological revolutions – e.g., the Second Industrial Revolution (displacement of skilled craftsmen and other manual jobs alongside the creation of assembly-line factory jobs, electricians, mechanics, construction workers, technicians, white-collar support jobs, etc.) and the Digital Revolution (displacement of routine clerical jobs alongside the creation of software development jobs, social media managers and content creators, call-center representatives, Uber drivers, etc.).
And even more significantly, the huge growth in service-sector jobs (hotels, leisure activities, transportation, restaurants, healthcare, education, retail, physicians, lawyers, etc.) that result from the economic growth and higher living standards that accompany technological revolutions. Economist Enrico Moretti has written about this extensively, arguing that in many economies with significant tech-related employment, five service sector jobs are often created for every new tech-related job. This has certainly been in the case in major technology hubs.
Some of you may have heard the oft-cited study by economist James Bessen about the development of ATMs and the predicted loss of bank teller jobs starting in the 1970s. The argument was very similar to what the doomers of their time have said about every major technological development – machines will replace workers, leading to high unemployment.
In fact, as the number of ATMs exploded, the number of bank teller jobs actually increased. Why? Well, it turns out the ATMs automated a bunch of routinized tasks previously performed by tellers, leading to a need for fewer bank teller per retail branch (i.e., the cost of operating a branch fell). But, as a result, bank executives decided to expand into new locations and open many more branches, driving up bank teller employment overall.
So, are we in the age of ATMs or the age of the English textile workers?
The real answer first – of course, we don’t know for certain.
But what we do know is that historical technological revolutions have delivered the good news we have seen above – significant net new job creation that ushered in eras of economic growth, productivity increases, and consumer surplus. So, to conclude otherwise, we would need to conclude that, “This time is different.”
Is it?
Some argue what’s different this time is the pace of technological change. AI is moving faster than previous revolutions; therefore, job displacement will happen way faster than new job creation and unemployment will spike. And, while previous technology advancements impacted more manual, clerical tasks, AI may also impact white collar jobs, creating even more potential job loss. Maybe.
But, if AI is that good it can replace lots of jobs very quickly, then that also means the organizations that eliminate those jobs must be experiencing enormous productivity growth. And we know that productivity growth means economic growth, which in turn means these same organizations can expand into new areas, build new products, and otherwise grow in ways were previously unprofitable. And new jobs will be created to support these growth initiatives and to support the new use cases AI enables – the new “social media manager” jobs of the AI era that previously didn’t exist.
Imagine the economic growth and consumer surplus possibilities if we can build things that were previously impossible – financially or technologically – as a result of AI. Imagine the advances in medicine that may dramatically change how well (and for how long) we live our lives. Imagine the consumer surplus that might be created by companies producing a whole new set of products we can’t even conceive of today. Imagine a world in which autonomous vehicles change the way we travel, expand the places we can live (because the nature of commuting is completely rethought), and change the nature of how cities are built (because the nearly 25% of urban space that has been dedicated to parking garages can be freed up for higher value uses). I could go on but will stop for now.
I was listening to a podcast recently with Ben Sasse, former senator from Nebraska and president of the University of Florida. The bulk of the podcast was about his current struggle with terminal cancer, but I thought his articulation of the potential for AI was pretty on point.
Sasse said, “Economics used to be a discipline about scarcity, but now it will be about ubiquitous abundance.” For most of our history, “scarcity” – whether of natural resources, financial resources, or even human ingenuity – has limited our ability to achieve “abundance.” But, if AI drives down the marginal cost of producing goods and services close to zero and enables innovation that has been beyond the imagination of even the most skilled innovators, the world that Sasse envisions will be pretty interesting.
Well, what is one to do?
For my OPM colleagues, I think that in times of change the best thing to do is to control what you can. We can’t turn back the clock on AI and wish it away. In fact, the AI that we know today is an “overnight success” 80+ years in the making. The first paper on neural networks – the theory human brain function can be modeled with computer operations and the foundation of much of today’s AI work – dates back to 1943!
And, by the way, along this 83-year path, we’ve had many “AI winters,” seeming breakthroughs at the time that created a lot of excitement that “this time was different” but ultimately culminating in hype over reality. Maybe the 2017 Google transformer paper that launched the most recent phase of AI excitement is the canary in the coal mine for the next AI winter of disappointment. Maybe – but hope is probably not your best strategy.
Instead, let’s control what we can.
Educate ourselves on how to utilize the technology to improve our daily job functions and deliver more efficiently on our promises to the American people. This doesn’t mean you have to transform your job but do look for incremental ways to try to new things. Take a training course, participate as an AI Ambassador, watch a YouTube video on some cool feature of AI. In my humble opinion, I think the greatest gift you can give yourselves is the opportunity for education – to learn new things that will enhance your skills and enable you to be a first-class participant in an AI-first world.
Think about what a world of “ubiquitous abundance” would mean in terms of our ability to serve our customers. What things are we not doing today because we don’t have the budget that we would want to do if budget were not a constraint? What incremental value could we bring to our annuitants, our health insurance customers, our partner agencies across the government if AI enabled us to free up 5%, 10%, 20% of work hours and add new things to our place of deliverables? I know many of you have great ideas; now is the time to start thinking “what if?”
Against this appeal for optimism, I also understand the anxiety some of you have. But I firmly believe we as Americans are fountains of ingenuity and innovation. And as a result, there will be all kinds of new roles that will emerge as we figure out how to expand the scope of things we do for our constituents.
Ultimately, each of you will decide your own truth. You don’t need to be an unfettered techno-optimist, but I encourage you not to be a Luddite either. As is often the case, the truth lies probably somewhere in between.