Editor’s note: We’re trying something new here. I read this back-and-forth exchange and decided we should give it a try. Or, more accurately, since we’re already having plenty of back-and-forth email exchanges like that, we just need to start publishing them. My friend Doug Hill agreed to be a guinea pig and chat with me about a subject that’s on both of our minds and publish it here.
Stogdill: I got this tweet over the holidays while I was reading your book. I mean, I literally got distracted by this tweet while I was reading your book:
— Ari Gesher (@alephbass) January 3, 2014
It felt like a natural moment of irony that I had to share with you. The author obviously has an optimistic point of view, and his claims seem non-obvious in light of our jobless recovery. I also recently read George Packer’s The Unwinding, and frankly it seems to better reflect the truth on the ground, at least if you get outside of the big five metro areas. But I suspect not a lot of techno optimists are spending time in places that won’t get 4G LTE for another year or two.
I’m not going to ask you what you think of the article because I think I already know the answer. I do have a few things on my mind, though. Is our jobless recovery a new structural reality brought about by more and more pervasive automation? Are Norbert Wiener’s predictions from the late 1940s finally coming true? Or is creative destruction still working, but just taking some time to adjust this time around? And, if one is skeptical of technology, is it like being skeptical of tectonics? You can’t change it, so bolt your house down?
Hill: Your timing is good. The day your email arrived the lead story in the news was the latest federal jobs report, which told us that the jobless “recovery” continues apace. Jobless, that is.
The national conversation about the impact of automation on employment continues apace, too. Thomas Friedman devoted his New York Times column a couple of days ago to The Second Machine Age, the new book by Erik Brynjolfsson and Andrew McAfee. They’re the professors from MIT whose previous book, Race Against the Machine, helped start that national conversation, in part because it demonstrated both an appreciation of automation’s advantages and an awareness that lots of workers could be left behind.
I’ll try to briefly answer your questions, Jim, and then note a couple points that strike me as somewhat incongruous in the current discussion.
Is our jobless economy a new structural reality brought about by more and more pervasive automation?
Yes. Traditional economic theory holds that advances in technology create jobs rather than eliminate them. Even if that maxim were true in the past (and it’s not universally accepted), many economists believe the pace of innovation in automation today is overturning it. This puts automation’s most fervent boosters in the odd position of arguing that technological advance will disrupt pretty much everything except traditional economic theory.
Are Norbert Wiener’s predictions from the late 1940s finally coming true? Or is creative destruction still working, but just taking some time to adjust this time around?
Yes to both. Wiener’s predictions that automation would be used to undermine labor are coming true, and creative destruction is still at work. The problem is that we won’t necessarily like what the destruction creates.
Now, about those incongruous points that bug me:
- First, a quibble over semantics. It’s convenient in our discussions about automation to use the word “robots,” but also misleading. Much, if not most, of the jobs displacement we’re seeing now is coming from systems and techniques that are facilitated by computers but less mechanical than the robots we typically envision assembling parts in factories. I don’t doubt that actual robots will be an ever-more-important force in the future, but they’ll be adding momentum to methods that corporations have been using for quite awhile now to increase productivity, even as they’re reducing payrolls.
- It’s commonly said that the answer to joblessness is education. Our employment problems will be solved by training people to do the sorts of jobs that the economy of the future will require. But wait a minute. If it’s true that the economy of the future will increasingly depend on automation, won’t we simply be educating people to do the sorts of jobs that eliminate more jobs?
- Techno optimists argue that our current employment problems are merely manifestations of a transition period on the way to a glorious future. “Let the robots take the jobs,” says Kevin Kelly, ”and let them help us dream up new work that matters.”
Even on his own terms, the future Kelly envisions seems more nightmarish than dreamlike. Everyone agrees automation is going to grow consistently more capable. As it does, Kelly says, robots will take over every job, including whatever new jobs we dream up to replace the previous jobs we lost to robots. If he’s right, we won’t be dreaming up new work that matters because we want to, but because we’ll have no choice. It will indeed be a race against the machines, and machines don’t get tired.
One more thing. You asked if being skeptical of technology is like being skeptical of tectonics. My first thought was to wonder whether anybody is really skeptical of tectonics, but given the polls on global warming, I guess anything is possible — more possible, I think, than a reversal of the robot revolution. So yeah, go ahead and bolt the house down.
Stogdill: Let me think where to start. My problem in this conversation is that I find myself arguing both sides of the question in my head, which makes it hard to present a coherent argument to you.
First, let me just say that I enter this discussion with some natural inclination toward a Schumpeterian point of view. In 1996, I visited a Ford electronics plant in Pennsylvania that was going through its own automation transformation. They had recently equipped the plant with then-new surface mount soldiering robots and redesigned the electronic modules that they produced there to take advantage of the tech. The remaining workers each tended two robots instead of placing parts on the boards themselves.
Except for this one guy. For some reason I’ve long forgotten, one of the boards they manufactured still required a single through-board capacitor, and a worker at that station placed capacitors in holes all day. Every 10 seconds for eight hours, a board would arrive in front of him, he would drop a capacitor’s legs through two little holes, push it over a bit to make sure it was all the way through, and then it was off to the next station to be soldiered. It was like watching Lucy in the Chocolate Factory.
I was horrified — but when I talked to him, he was bound and determined to keep that job from being automated. I simply couldn’t understand it. Why wouldn’t he want to upgrade his skills and run one of the more complex machines? Even now, when I’m more sympathetic to his plight, I’m still mystified that he could stand to continue doing that job when something else might have been available.
Yet, these days I find myself losing patience with the reflexive trope “of course technology creates more jobs than it destroys; it always has. What are you, a Luddite?” The Earth will keep rotating around the sun, too; it always has, right up until the sun supernovas, and then it won’t.
Which isn’t to say that the robots are about to supernova, but that arguments that depend on the past perpetuating into the future are not arguments — they’re wishes. (See also, China’s economy will always keep growing at a torrid rate despite over-reliance on investment at the expense of consumption because it always has). And I just can’t really take anyone seriously who makes an argument like that if they can’t explain the mechanisms that will continue to make it true.
So, to my thinking, this boils down to a few key questions. Was that argument even really true in the past, at least the recent past? If it was, is the present enough like the past that we can assume that, with re-training, we’ll find work for the people being displaced by this round of automation? Or, is it possible that something structurally different is happening now? And, even if we still believe in the creative part of creative destruction, what destructive pace can our society absorb and are there policies that we should be enacting to manage and ease the transition?
This article does a nice job of explaining what I think might be different this time with its description of the “cognitive elite.” As automation takes the next layer of jobs at the current bottom, we humans are asked to do more and more complex stuff, higher up the value hierarchy. But what if we can’t? Or, if not enough of us can? What if it’s not a matter of just retraining — what if we’re just not talented enough? The result would surely be a supply/demand mismatch at the high end of the cognitive scale, and we’d expect a dumbbell shape to develop in our income distribution curve. Or, in other words, we’d expect new Stanford grads going to Google to make $100K and everyone else to work at Walmart. And more and more, that seems like it’s happening.
Anyway, right now I’m all question, no answer. Others are suggesting that this jobless recovery has nothing to do with automation. It’s the (lack of) unions, stupid. I really don’t know, but I think we — meaning we technologists and engineers — need to be willing to ask the question “is something different this go-round?” and not just roll out the old history-is-future tropes.
We’re trying to create that conversation at least a bit by holding an Oxford-style debate at our next Strata conference. The statement we’ll be debating is: “Technology creates more jobs than it destroys,” and I’ll be doing my best to moderate in an even-handed way.
By the way, your point that it’s “not just robots” is well taken. I was talking to someone recently who works in the business process automation space, and they’ve begun to refer to those processes as “robots,” too — even though they have no physical manifestation. I was using the term in that broad sense, too.
Hill: In your last email you made two points in passing that I’d like to agree with right off the bat.
One is your comment that, when it comes to predicting what impact automation will have on employment, you find yourself “arguing both sides of the question.” Technology always has and always will cut both ways, so we can be reasonably certain that, whatever happens, both sides of the question are going to come into play. That’s about the only certainty we have, really, which is why I also liked it when you said, “I’m all question, no answer.” That’s true of all of us, whether we admit it or not.
We are obligated, nonetheless, to take what Norbert Wiener called “the imaginative forward glance.” For what it’s worth then, my answer to your question, “Is something different this go-round?” — by which you meant, even if it was once true that technological advancement created more rather than fewer jobs, that may no longer be true, given the pace, scale, and scope of the advances in automation we’re witnessing today — is yes and no.
That is, yes, I do think the scope and scale of technological change we’re seeing today presents us with challenges of a different order of magnitude than what we’ve faced previously. At the same time, I think it’s also true that we can look to the past to gain some sense of where automation might be taking us in the future.
In the articles on this issue you and I have traded back and forth over the past several weeks, I notice that two of the most optimistic, as far as our automated future is concerned, ran in the Washington Post. I want to go on record as denying any suspicion that Jeff Bezos had anything to do with that.
Still, the most recent of those articles, James Bessen’s piece on the lessons to be learned from the experience of America’s first industrial-scale textile factories (“Will robots steal our jobs? The humble loom suggests not”) was so confidently upbeat that I’m sure Bezos would have approved. It may be useful, for that reason, to take a closer look at some of Bessen’s claims.
To hear him tell it, the early mills in Waltham and Lowell, Massachusetts, were 19th-century precursors of the cushy working conditions enjoyed in Silicon Valley today. The mill owners recruited educated, middle-class young women from surrounding farm communities and supplied them with places to live, houses of worship, a lecture hall, a library, a savings bank, and a hospital.
“Lowell marked a bold social experiment,” Bessen says, “for a society where, not so long before, the activity of young, unmarried women had been circumscribed by the Puritan establishment.”
The suggestion that the Lowell mills were somehow responsible for liberating young women from the clutches of Puritanism is questionable — the power of the Puritan church had been dissipating for all sorts of reasons for more than a century before factories appeared on the banks of the Merrimack — but let that go.
It is true that, in the beginning, the mills offered young women from middle-class families an unprecedented opportunity for a taste of freedom before they married and settled down. Because their parents were relatively secure financially, they could afford to leave them temporarily behind without leaving them destitute. That’s a long way from saying that the mills represented some beneficent “social experiment” in which management took a special interest in cultivating the well-being of the women they employed.
Thomas Dublin’s Women at Work: The Transformation of Work and Community in Lowell, Massachusetts, 1826-1860 tells a different story. Women were recruited to staff the mills, Dublin says, because they were an available source of labor (men were working the farms or employed in smaller-scale factories in the cities) and because they could be paid less than men. All supervisory positions in the mills were held by men. Also contrary to Bessen’s contention, the women weren’t hired because they were smart enough to learn specialized skills. Women tended the machines; they didn’t run them. “To the extent that jobs did not require special training, strength or endurance, or expose operatives to the risk of injury,” Dublin says, “women were employed.”
How much time they had to enjoy the amenities supposedly provided by management is another question. According to Dublin, mill workers put in 12 hours a day, six days a week, with only three regular holidays a year. As the number of mills increased, so did the pressure to make laborers more productive. Speedups and stretch-outs were imposed. A speedup meant the pace of the machinery was increased, a stretch-out meant that each employee was required to tend additional pieces of machinery. Periodic cuts in piece wages were another fact of mill life.
Because of their middle-class backgrounds, and because they were accustomed to pre-industrial standards of propriety, the first generation of women felt empowered enough to protest these conditions, to little avail. Management offered few concessions, and many women left. The generation of women who replaced them were less likely to protest. Most had fled the Irish famine and had no middle-class homes to return to.
I go into this in some detail, Jim, because it’s important to acknowledge what automation’s fundamental purpose has always been: to increase management profits. Bold social experiments to benefit workers haven’t figured prominently in the equation.
It’s true that factory jobs have, in the long run, raised the standard of living for millions of workers (the guy you met in the Ford electronics plant comes to mind), but we shouldn’t kid ourselves that they’ve necessarily been pleasant, fulfilling ways to make a living. Nor should we kid ourselves that management won’t use automation to eliminate jobs in the future, if automation offers opportunities to increase profits.
We also need to consider whether basing our economy on the production and sale of ever-higher piles of consumables, however they’re manufactured, is a model the planet can sustain any longer. That’s the essential dilemma we face, I think. We must have jobs, but they have to be directed toward some other purpose.
I realize I haven’t addressed, at least directly, any of the questions posed in your email. Sorry about that — the Bessen article got under my skin.
Stogdill: That Bessen article did get under your skin, didn’t it? Well, anger in the face of ill-considered certainty is reasonable as far as I’m concerned. Unearned certainty strikes me as the disease of our age.
Reading your response, I had a whole swirl of things running through my head. Starting with, “Does human dignity require meaningful employment?” I mean, separate from the economic considerations, what if we’re just wired to be happier when we grow or hunt for our own food? — and will the abstractions necessary to thrive in an automation economy satisfy those needs?
Also, with regard to your comments about how much stuff do we need — does that question even really matter? Is perpetual sustainability on a planet where the Second Law of Thermodynamics holds sway even possible? Anyway, that diversion can wait for another day.
Let me just close this exchange by focusing for just one moment on your point that productivity gains have always been about increasing management profits. Of course they have. I don’t think that has ever been in question. Productivity gains are where every increase in wealth ever has come from (except for that first moment when someone stumbled on something useful bubbling out of the ground), and profit is how is how we incent investment in productivity. The question is how widely gains will be shared.
Historically, that argument has been about the mechanisms (and politics) to appropriately distribute productivity gains between capital and labor. That was the fundamental argument of the 20th century, and we fought and died over it — and for maybe 30 years, reached maybe a reasonable answer.
But what if we are automating to a point where there will be no meaningful link between labor and capital? There will still be labor, of course, but it will be doing these abstract “high value” things that have nothing whatsoever to do with the bottom three layers of Maslov’s hierarchy. In a world without labor directly tied to capital and its productivity gains, can we expect the mechanisms of the 20th century to have any impact at all? Can we even imagine a mechanism that flows the value produced by robots to the humans they sidelined? Can unemployed people join a union?
We didn’t answer these questions, but thanks for exploring them with me.