FIRST OF ALL, THIS IS AN EXTREMELY IMPORTANT BOOK
Race Against the Machine deserves praise for jump-starting an important discussion about the effect of technology on our economy. As the authors point out, the impact of computers and information technology has been largely left out of most analysis regarding causes of our current unemployment woes. This book, therefore, is an attempt to “put technology back in the discussion.”
BUT IT SUFFERS FROM SCHIZOPHRENIA
The book is divided into roughly two halves: one pessimistic and one optimistic. The first three chapters comprise the pessimistic portion and make a compelling case for how accelerating technological progress and rapid productivity gains are not only creating unemployment, but also contributing to greater inequality. The last two chapters take a more optimistic tone and attempt to lay out possible solutions to these problems.
Unfortunately, the pessimistic chapters are much more convincing. The net effect is rather dissonant, not unlike a general rounding up his troops and announcing, “Gentlemen, we’re outmatched in every possible way. Now go out there and win!”
A QUICK SUMMARY OF THE “PESSIMISTIC” CHAPTERS
Race Against the Machine begins by building a strong argument for technological unemployment. The authors draw our attention to recent innovations like self-driving cars and Jeopardy-winning computers and show how such advances threaten to encroach further and further into realms once dominated by human labor. Such technologies ultimately affect all sectors of the economy since computers are a prime example of what economists call a “General Purpose Technology.” Like steam power and electricity, computers reside in a category of innovation “so powerful that they interrupt and accelerate the normal march of economic progress.” Moreover, thanks to Moore’s law, computers are improving at an exponential rate, so we can expect very disruptive changes to arrive very quickly indeed.
Making matters worse, a host of related trends result in the benefits of technological progress not being shared equally. Not only is the value of highly skilled workers diverging sharply from that of low skilled workers, but the value of capital is increasing relative to that of labor. Also important is the “superstar effect.” Technological advances extend the reach of superstars in various fields and allow them to win ever larger market shares. This comes at the expense of the many, since “good, but not great, local competitors are increasingly crowded out of their markets.” Putting all this together you get the central thesis of Race Against the Machine: namely, that digital technologies are outpacing the ability of our skills and organizations to adapt.
HOW THE AUTHORS DISTANCE THEMSELVES FROM THE “END-OF-WORK”
In some ways this is a familiar argument. The idea that technology might replace the need for human workers on a large scale has been floating around for a long time, at least since the industrial revolution. However, it should be noted that early on the authors distance themselves from what they call the “end-of-work” crowd, thinkers like Jeremy Rifkin, Martin Ford, and even John Maynard Keynes, who have argued that with time the role of human labor is bound to diminish. Certainly, one can understand why the authors would be cagey about being associated with this idea. Historically, predictions regarding the coming obsolescence of human labor have been wildly exaggerated, and economists generally view such arguments as fallacious.
So what differentiates Race Against the Machine from a more traditional end-of-work argument? According to the authors:
“So we agree with the end-of-work crowd that computerization is bringing deep changes, but we’re not as pessimistic as they are. We don’t believe in the coming obsolescence of all human workers. In fact, some human skills are more valuable than ever, even in an age of incredibly powerful and capable digital technologies. But other skills have become worthless, and people who hold the wrong ones now find that they have little to offer employers. They’re losing the race against the machine, a fact reflected in today’s employment statistics.”
In short, the authors are more optimistic because they believe there will still be plenty of jobs for humans in the future. We just need to update our skills and organizations to cope with new digital technologies, and then we will be able to create new avenues of employment and save our struggling economy.
BUT THE “OPTIMISTIC” CHAPTERS AREN’T SO ENCOURAGING
Which brings us to those last two chapters, the unconvincing ones I alluded to earlier. For I believe Race Against the Machine suffers from the same problem as a lot of nonfiction books: It does a great job of stating the problem, but a not-so-great job of laying out the solution.
SO HOW ABOUT WE RACE WITH MACHINES?
The first suggestion the authors make can be summarized as “race with machines.” A human-machine combo has the potential to be much more powerful than either a human or machine alone. So therefore it’s not simply a question of machines replacing humans. It’s a question of how can humans and machines best work together.
I don’t disagree with this point on the surface. But I fail to see how it suggests a way out of our current predicament. The human-machine combo is a major cause of the superstar economics described earlier in the book. Strengthen the human-machine combo and the superstar effect will only get worse. In addition, if computers are encroaching further and further into the world of human skills, won’t the percentage of human in the human-machine partnership just keep shrinking? And at an exponential pace?
Moreover, as I’ve written about before on this site, the human-machine partnership can sometimes be less than the sum of its parts. Consider the example of airline pilots:
“In a draft report cited by the Associated Press in July, the agency stated that pilots sometimes “abdicate too much responsibility to automated systems.” Automation encumbers pilots with too much help, and at some point the babysitter becomes the baby, hindering the software rather than helping it. This is the problem of “de-skilling,” and it is an argument for either using humans alone, or machines alone, but not putting them together.” (link)
PERHAPS ORGANIZATIONAL INNOVATION WILL SAVE US
The authors go on to discuss the importance of “organizational innovation.” In particular, they discuss the creation of new business platforms that might empower humans to compete in new marketplaces.
Again, I agree in theory. Certainly some new platform may hold the key to productively mobilizing the unemployed. But current examples are far from encouraging. The authors cite websites like eBay, Apple’s App Store, and Threadless. An obvious point would be that the kind of people who are able to hustle and make a living on such websites are not exactly average workers in any sense of the word. Not everyone can run an online retail store, program an app, or design a t-shirt. But that’s beside the point. The question we should be asking is will such online marketplaces grow in the future? Perhaps they will expand to the point that they can encompass more and more ordinary workers?
I am highly skeptical. Once again, superstar economics apply here, since effectively everyone in these markets is competing with everyone else. One potential solution is the growth of niche markets. If you focus on selling unique items to a niche audience perhaps you can carve out your own little market in which you are the lone superstar.
But this idea also has its problems. How many niches can there possibly be? Enough to provide employment for the legions of truck drivers and supermarket checkers who may soon be exiting the workforce?
When discussing technology and unemployment, I think it is important not to leave digital abundance out of the discussion. Digital abundance has the potential to be just as disruptive as automation. Traditional businesses are under attack from two sides. Services are being automated, while at the same time goods are being digitized.
Imagine the domestic entrepreneur who has started his own eBay store. He sells niche action figures to a few enthusiastic fans. Nonetheless, enough fans exist that he can make a decent living, all thanks to the wonders of eBay’s “organizational innovation.”
Enter affordable desktop 3D printing, a technology that is rapidly arriving on the scene. All of a sudden once eager customers can buy cheap raw materials and print all the action figures they want. Digital files containing the precise specs for figures get designed, released, and traded extensively on file sharing sites. An explosion of innovation for sure, but also a potential threat to a business model that focuses too much on the sale of unique tangible goods.
Thought experiments like this reveal why intellectual property and digital rights management are going to become increasingly hotly debated issues. As tangible goods become digitized they go from being tangible property to intellectual property. So the efficacy of a lot of future businesses depends on the efficacy of intellectual property, and a survey of recent history quickly reveals the troubles inherent in this area.
OKAY, BUT AREN’T THERE OTHER TYPES OF ORGANIZATIONAL INNOVATION?
So far I have focused on marketplaces for goods. I should note that there are also online labor marketplaces like Taskrabbit and Mechanical Turk. These websites provide a great service by efficiently matching demand for labor to humans willing to work. While increasing efficiency is beneficial, such websites will be of limited help if demand for average-skilled labor falls in the aggregate.
Now I don’t want to sound overly pessimistic. In general, I would agree that the unemployed represent a huge slack resource, and quite possibly somebody is going to come up with some previously unimagined way to harness this large pool of people. But at the moment, such organizational innovation is just a theory. I do not see the seeds of a workable solution in the current crop of platforms.
As their final example of organizational innovation, the authors mention the promise of “micromultinationals.” They write:
“Technology enables more and more opportunities for what Google chief economist Hal Varian calls “micromultinationals”—businesses with less than a dozen employees that sell to customers worldwide and often draw on supplier and partner networks. While the archetypal 20th-century multinational was one of a small number of megafirms with huge fixed costs and thousands of employees, the coming century will give birth to thousands of small multinationals with low fixed costs and a small number of employees each.”
I don’t know if this quote is meant to be taken literally, but for fun let’s crunch some numbers. The coming century (100 years) will give birth to thousands (max: 9999) multinationals with low numbers of employees (less than 12). Therefore:
9999 x 11 / 100 = 1100 jobs/year. Not exactly encouraging.
LOOKING TO COMBINATORIAL EXPLOSION FOR HOPE
Early on in the book, the authors take time to explain the incredible power of exponential growth. They discuss Moore’s law and quote Ray Kurzweil’s book The Singularity is Near to illustrate how exponential growth can quickly shift from modest gains to “jaw-dropping” changes.
This puts us in a dire place. If we accept the authors’ premise of a losing race, in which technology (progressing exponentially) is outrunning our skills and our institutions, then how can society hope to catch up? Trying to win a race against exponential growth sounds like an impossible task.
In chapter four, the authors claim to come up with the answer: “combinatorial explosion.”
Combinatorial explosion is the idea that new ideas are combinations of two or more old ideas. Since digital technologies facilitate the easy exchange of information, and ideas—unlike physical resources—can’t be used up, we therefore have virtually limitless possibilities for innovation.
“Combinatorial explosion is one of the few mathematical functions that outgrows an exponential trend. And that means that combinatorial innovation is the best way for human ingenuity to stay in the race with Moore’s Law.”
This suggests a strange dichotomy. As if Moore’s Law is the exclusive tool of machines, while combinatorial explosion is the exclusive tool of humans. This is clearly false. Combinatorial explosion is a huge cause of our current situation. It is a primary reason why disruptive technologies are moving so fast in the first place.
Here’s an easy example: IBM is re-purposing Watson—the Jeopardy-winning computer—to perform medical diagnosis. So here we see one idea colliding with another in true combinatorial fashion, and what’s the result? Yet another potential threat to jobs, this time in the medical field.
HYPERSPECIALIZATION AND INFINITE NUMBERS OF MARKETS
Hyperspecialization is the authors’ answer to the problem of superstar economics:
“In principle, tens of millions of people could each be a leading performer—even the top expert—in tens of millions of distinct, value-creating fields.”
In principle maybe. In practice there are huge obstacles. Again, what’s to stop one superstar-machine combo from just dominating multiple fields? Or even just one machine? In the health care industry for example, computers like Watson are going to be able to mine the literature of all fields of medicine. After all, to a computer, what’s a few thousand more documents to read? Machines can rapidly scale up their expertise in ways that humans simply can’t.
More importantly we should examine the term “value-creating fields.” Value under our current system is closely tied to scarcity. Digital abundance directly undermines this source of value. So once again we are confronted with an intellectual property challenge. If we are going to have an economy where everyone is an expert in a different field and produces “bits,” we are going to need a mechanism by which these non-scarce bits translate into an income. The truth is we already have numerous such experts. The Internet is overflowing with amateurs who voluntarily immerse themselves in hyper-specialized tasks purely for enjoyment, not because this path is necessarily a viable strategy for making money.
BUT DON’T WE ALL HAVE SOMETHING UNIQUE TO CONTRIBUTE?
Yes of course. Human creativity is astounding, and everyone has something to offer. But peoples’ output—however unique, interesting, and valuable—will not necessarily be monetizable. Especially in an abundant digital environment.
At this point, I want to return to a stray quote from earlier in the book, because I think it presents the opportunity to make an important point.
“…digital technologies create enormous opportunities for individuals to use their unique and dispersed knowledge for the benefit of the whole economy.”
When I look at the Internet and communications technologies, I see a huge threat to this “dispersed knowledge.” The Internet has a way of destroying information asymmetry, which is another important factor to consider when looking at future employment. Any jobs that depend upon having exclusive access to knowledge that no one else has are potentially at risk in a world where increasingly everyone is connected and data is widely shared.
INVESTMENT IN HUMAN CAPITAL
Education seems like the most straightforward solution to our problem. If our skills are falling behind, then we’d better acquire new skills right?
I am highly dubious of education’s ability to solve this problem. For one, the most promising experiments in education right now, those that use technology aggressively, like Khan Academy or Stanford’s online courses, have the potential to create unemployment in the education field. In addition to the long-term promise of fully automated learning environments, superstar economics rears its head once again. After all, Khan Academy is built around a superstar teacher: Khan himself. And Stanford’s recent Artificial Intelligence course allowed one professor to effectively reach 58,000 students.
The authors do present an important counter:
“Local human teachers, tutors, and peer tutoring can easily be incorporated into the system to provide some of the kinds of value that the technology can’t do well, such as emotional support and less-structured instruction and assessment. For instance, creative writing, arts instruction, and other “soft skills” are not always as amenable to rule-based software or distance learning.”
Person-to-person interaction is indeed an important aspect in a lot of teaching, and won’t be vanishing any time soon. However, it arguably becomes less important the further you move up the educational ladder. A fifth grader needs lots of emotional support and hands-on instruction, but a self-directed higher education student may need almost none. College is so expensive right now that I can imagine people increasingly forgoing it in favor of cheaper, more automated learning options.
But even with person-to-person learning, technological advances mean that single tutors or teachers will increasingly be able to meet the needs of more and more students. If the learning software does a halfway decent job, then the necessity of human intervention should decrease with time.
In addition, going back to the idea of digital abundance, such human intervention may be increasingly available for free. Already, it’s stunningly easy to go on the Internet and find volunteers who will provide emotional support and helpful feedback, at zero cost. Online communities around “soft skills” like creative writing are particularly vibrant, and offer the opportunity to develop a craft with a huge support network that easily rivals what you would get from a traditional paid learning experience.
I think there is a cultural bias towards judging online interactions as somehow always less valuable than real space interactions. With every passing year, as the resolution of communications technologies increases, this point of view becomes increasingly absurd. Cultural norms may move slowly, but I suspect they will eventually come around on this issue.
But all of these considerations aside, there is a much bigger problem. One cannot escape the simple truth that humans learn slowly and technology advances quickly. If we take exponential growth seriously, how can education expect to keep up? Are we going to retrain unemployed truck drivers to become app programmers? Chances are by the time such retraining is complete, technology will have moved on. And don’t forget that the machines themselves will increasingly be educating themselves.
One solution might be augmenting human thinking capability. If we could upgrade humans the way we upgrade machines, then “the race” would be over, and racing with machines would make more sense. This may sound far-fetched, but in these futuristic times nothing should be ruled out. In Andrew McAfee’s own words, “Never say never about technology.” The question is when will such technologies arrive? And what societal upheaval might we be in for in the meantime?
The authors make a series of common sense policy recommendations that affect institutions like education, business, and law. Most of these suggestions are great, and might lead to a better society, but it is unclear how any of them will create jobs. Rather the goal of these suggestions seems to be to allow innovation and progress to flourish, which in my opinion may only accelerate the process of job loss, as per my arguments above.
The one exception might be suggestion number 13:
“Make it comparatively more attractive to hire a person than to buy more technology. This can be done by, among other things, decreasing employer payroll taxes and providing subsidies or tax breaks for employing people who have been out of work for a long time. Taxes on congestion and pollution can more than make up for reduced labor taxes.”
I suppose this might help keep some jobs around longer, but at the expense of investment in technology that presumably we would want to encourage. This seems directly at odds with the authors’ other more pro-innovation suggestions. Do we want technological progress or not?
RE-EVALUATING THE DESIRABILITY OF “WORK”
One solution the authors quickly dismiss is wealth redistribution. Their reasoning?
“While redistribution ameliorates the material costs of inequality, and that’s not a bad thing, it doesn’t address the root of the problems our economy is facing. By itself, redistribution does nothing to make unemployed workers productive again. Furthermore, the value of gainful work is far more than the money earned. There is also the psychological value that almost all people place on doing something useful. Forced idleness is not the same as voluntary leisure.”
I think a distinction ought to be made between wage labor and other perfectly meaningful ways of occupying one’s time. Looking ahead, perhaps our cultural reverence for wage labor is misplaced. After all, one way to look at wage labor is it is a mechanism which forces us to spend time on what the short-term market thinks is valuable, rather than on what we as individuals think is valuable. Sure, lots of people love their jobs. But lots of people hate their jobs too. If you liberated all the people who hate their jobs from the constraints of wage labor, chances are a decent portion of them might find something else more productive to spend their time on. And society as a whole might benefit tremendously.
I do not rule out the possibility that eventually we may find a way to gainfully employ everybody while keeping our current system intact. If some human innovator does not crack this problem, then certainly a sufficiently powerful computer might. But I shouldn’t have to point out the absurdity of asking in effect, “Hey Hal, can you figure out what jobs we should be doing?” It seems to me that long before then, we ought to be re-examining whether we even want traditional jobs in the first place. Perhaps we should be working with machines in order to win the race against labor.
The authors repeatedly state that our institutions are losing a race with technology. But they do not consider the possibility that our economy itself might be one of these trailing institutions.
NOT TO OVERSTATE THE LEVEL OF DISAGREEMENT
Near the end of the book, the authors do admit that there are “limits to organizational innovation and human capital investment.” So most likely they would not be too surprised by many of the criticisms I’ve leveled above.
And I would agree with the authors that, excepting the jobs issue, all of these new technologies are unequivocally a good thing. There is clearly a lot of cause for optimism in the general sense.
In addition, if we can advance technology quickly enough there are two long term solutions to the unemployment problem. The first, direct intelligence augmentation, I have already mentioned above. The second solution involves cost of living. Specifically, if technology can significantly lower costs of living, then declining income prospects for average individuals will not sting so much. However, to accomplish this we would have to see dramatic drops in the price of essential goods like housing and healthcare, drops which may happen eventually, but may not arrive quickly enough to prevent social unrest.
In the final assessment, I think my biggest issue with this book is the way authors fail to effectively distinguish themselves from the “end-of-work” crowd. After stating in the opening chapter that they are “more optimistic” about the future of human labor, they do not present any credible reasons for such optimism. The authors may claim they do not believe in the “end of work,” but their claims will not prevent me from filing them next to Martin Ford and Jeremy Rifkin on my bookshelf.