Average Is Over: Powering America Beyond the Age of the Great Stagnation (Tyler Cowen)

Tyler Cowen is a popular economist, known for an influential blog (Marginal Revolution) and a set of books on economics directed at a general audience.  In Average Is Over, a book from 2013, Cowen predicts an American future of increased economic (and thus social) division, as new technology enables those most conversant with it to profit, and forces others to be paid less as they become relatively less productive.  This is a common historical occurrence, of course, where those whose skills are no longer valued by the market, from hand weavers to buggy whip makers to floppy disk craftsmen, must ultimately retire or retool, often never regaining their previous income, though the economy as a whole, and thus average and median income, expand in the long run despite short- and medium-term pain and dislocations.  The difference in Cowen’s analysis is that he forecasts a permanent division, the result of ever-improving radically new technology and the failure of some, or many, to properly orient themselves with respect to that technology.

Maybe.  But Cowen’s analysis is flawed by tunnel vision and an obtuse, reflexive belief in the progress of technology.  All of Cowen’s points in this book revolve around a single premise—that the abilities of machine intelligence will continue to increase at an exponential rate, and that how well each human can leverage such machine intelligence in his personal productivity will determine his personal economic fate.  Cowen does not believe that we will be ruled by our new machine overlords; he does not appear to predict “strong” artificial intelligence (i.e., machines exhibiting human cognition).  Rather, his focus is on “integration of capabilities” between man and machine, with the latter exhibiting “weak” artificial intelligence.  While this makes his vision less ludicrous than, say, Ray Kurzweil’s crypto-religious belief in the so-called Singularity (whose just-around-the-corner date is ever pushed into the future), Cowen offers little evidence other than wishful thinking that his own vision is coming true.

Cowen begins, in Part I, “Welcome to the Hyper-Meritocracy,” by discussing just what being an economic loser means in today’s world, or, to use a technical term, what “labor market polarization” means in practice.  Median real wages have been declining for decades, and continue to decline.  At the same time, the upper end of American jobs, in essence skilled labor, has seen rising wages.  Such skills can be technical, but are often instead marketing skills (valuable to obtain the attention of the upper classes, who have money to spend) and managerial skills (valuable to increase efficiency of teams needed for complex projects).  Moreover, those who have no such concrete skills, but do have the ability to perform general work in a precise, conscientious and disciplined manner, also see their wages and job prospects rise.  (Cowen treats as undisputed that more women than men have such “manner skills,” which is certainly true, though daring in that most writers today feel constrained to deny reality if it might impinge on political correctness, which, after all, is 80% pure denial of reality).  Ultimately, what matters most, and what allows (according to Cowen) ignorant 22-year-old consultants and finance majors to command large salaries, is high “general intelligence”—which can be used in a flexible manner as applied to different tasks, or, fitting Cowen’s focus, to mesh with technology to increase productivity.

It’s not just that median real wages continue to decline.  It’s that the jobs have gone away and they aren’t coming back, which results in the declining wages, but has no obvious or easy solution.  There are many “zero marginal product” workers (i.e., those who provide less value than the costs they impose, either by their own characteristics or because of external costs like regulation, healthcare and lawsuits), who will never be hired.  Freelancing or the “gig economy” isn’t a solution—it’s a way for people to work hard and get little, usually.  Kind of like Lagos.  Or like Berlin, where living is cheap and a lot of people don’t bother to work hard or consistently, because they don’t have to and the culture doesn’t even encourage it.  “In a wealthy society, sometimes it’s just enough to get by and have a good time.”  In other words, especially with government benefits, the less successful can not just merely subsist, as might be true in the past, but find lives that are much to their liking, while contributing nothing to economic productivity (and being parasites on the productive).

Little of this is explored in detail (nor is it original) since this is a short book, and the chapters read a bit too much like expanded blog posts.  On the other hand, Cowen has a gift for the pithy insight—for example, he notes that positive buzzwords offered by employers like Google, such as “teamwork,” “morale,” and “integrity,” only are relevant to that small number of people hired.  From the perspective of the many not hired, a different, equally valid, word characterizes Google’s hiring practices:  “exclusion.”  That is, “There is no high morale without exclusion, no integrity without exclusion, and no corporate culture without exclusion.”  Needless to say, Google doesn’t mention exclusion in its HR materials.  These pithy insights at least partially compensate for the lack of detail, by providing reader interest if not depth of analysis.

Part II, which is more than half the book, is titled “What Games Are Teaching Us.”  Here, Cowen introduces his obsession—chess computers, and how they supposedly illustrate our future.  I guess chess computers are modestly interesting.  But Cowen’s logical leaps based on chess computers are entirely unjustified on any evidence he offers.  He errs by believing that if something can be imagined, it is inevitable.  So, merely because chess computers can combine with human choice in an obscure and frenetic chess hybrid called “Freestyle,” Cowen believes it is inevitable computers will combine with humans to dictate negotiation of business deals and each individual’s moment-to-moment actions on dates.  “We’re going to generate a lot of hairy, very complicated personal interactions, driven by real-time data analysis and computer intelligence.”  No, we’re not, or at least Cowen offers no evidence that the tightly constrained chess model has any relevance whatsoever to any other area of human endeavor, action, or interaction.  And just because an algorithm can be created by Netflix that recommends movies based on other choices says nothing to the contrary.  There is a wide chasm between using machines to deliver facts or perform set operations, and using machines to address situations with inherent and irresolvable ambiguities (like pretty much everything in human life, and in particular human interactions).

In fact, machines can’t even do those basic, simple things well.  Cowen tells us that even now, “Whether it is through Siri, Google, or Wikipedia, there is now almost always a way to ask and—more importantly—a way to receive the answer in relatively digestible form.”  Even this mild claim is false.  The only answers one can receive in any of those ways are answers that are rule-bound and unambiguous, as with chess.  All of those technologies, for anything requiring subtlety, ambiguity or analysis, much less judgment, are essentially worthless (and in the case of Wikipedia especially, are often filled with falsehoods even for binary questions).  Just because most of the time the Internet will accurately tell us the time and location of a movie showing, tells us nothing about the likely future abilities of computers to assist in anything complex, ambiguous, or relying in any way on human intuition, judgment, emotions and impulses.

Thus, Cowen is far too optimistic about the near future abilities of machine “intelligence” (which is, of course, not intelligence at all, though Cowen refuses to admit it or even address the distinction).  He is careful to not be overtly dewy-eyed about technology; he notes, for example, that Apple’s Siri “disappoints with its mistakes and frequently obtuse responses,” but still thinks it, and its ilk, will “improve rapidly.”  Maybe they will.  And maybe they won’t.  Chess computers, Cowen’s only example of refined improvement (maybe because there is an ending point that can be identified and quantified) are extremely poor as a template for machine intelligence in the larger world, since chess has a very constrained, wholly artificial world, totally divorced from the real world and humanity, and operating within very clear rules.  In chess, there is no room for ambiguity and errors necessarily become evident, or are prevented from ever occurring by the existence of the rules.  Judgment, something that machines will never provide, is wholly unnecessary.  Cowen acknowledges that chess is not an exact analogue, but you can tell he really doesn’t think there’s any real difference between a chess computer and a machine that can tell you how to act moment-to-moment on a first date.  This is a blinkered vision that refuses to admit that human judgment cannot, or at least never will, be reduced to a set of measured inputs subjected to an algorithm.

Similarly, Cowen describes the very mixed bag of technology assistance existing with GPS systems, gasoline pumps, self-checkout lines and the rest, fully admitting that they are far from perfect, and often worse than the human-centered alternatives.  But then he announces that “this problem will someday be solved, and machines will solve it.”  Why?  We are not told.  He then doubles down, claiming that someday he will be able to “walk into the house and announce to the sensors:  ‘Cox cable company, I am home.  Some branches fell at my house and now the cable wire is hanging dangerously low against some bushes.  Please let me know some times when someone could come by to fix it.”  I doubt if this will ever happen.  But more to the point, Cowen offers us nothing but faith that it will, or that it will not be accompanied by the mandate to watch a glitchy two-minute advertising video before he is permitted to address the sensors.  For Cowen, the future will always be better, if it is centered around technology.  He thinks anything can be measured, objectively quantified and communicated.  He really thinks, he really does, that someday when evaluating a corporate lawyer, “Siri will tell you: ‘This lawyer’s written briefs are in the top eighty-first percentile of his peer group; that explains thirty-eight percent of performance on a corporate deal.”  This is beyond silly, it is stupid, because there is no way to measure such things, and there never will be (and it also shows ignorance—corporate deals do not involve briefs, ever, which are purely a litigation device).  Thus, when Cowen concludes that as a result of all these magic machines, the gap between the successful and the unsuccessful will grow, the reader chuckles grimly, because the magic machines are ludicrous.  The gap may grow, but this is not why.

This set of claims, by itself, would merely be somewhat silly and somewhat interesting (though we get way too much information about the details of Freestyle).  But this is the trope that Cowen then carries through the rest of the book—intelligent machines will combine with the most intelligent humans to produce results unavailable without using both machine and human, and those humans unable or unwilling to so act will be left behind, so many hand weavers and floppy disk craftsmen.

Cowen tries to bring it all together in Part III, “The New World of Work.”  While admitting that foreign outsourcing hurts some of us, he toes the standard “collectively we’re better off” line, and claims that “at a fundamental moral level a job for a ‘foreigner’ is every bit as worth an outcome as a job for a ‘real American.’”  Neoliberalism for all!  Nobody tell Trump, though, that his moral betters have found the real answer and it involves debasing Americans to favor foreigners.  Similarly, within the United States (and other countries), some geographic areas (certain cities, namely) will get rich, as the smart, machine-integrated and well-connected congregate there (what Cowen calls “clustering of commercial talent”), while others get even poorer and “hollow out.”  This is fine by Cowen.

It’s not that Cowen says there is no problem at all.  Cowen does offer some ways to lift everybody’s boat.  All are dubious.  For example, he pushes education, namely MOOCs, which since this book was written have in no way fulfilled their promise, or any promise (in part doubtless because college education today usually offers little education, but rather is mostly a filtering device used by employers, and a place for elites to meet other elites).  And then he goes off the rails.  He says computer games such as role-playing multiplayer games (like World of Warcraft) educate the young and describes the idea that a game designer should someday get the Nobel Peace Prize as “entirely reasonable.”  His proof of such education by games is that chess prodigies are getting younger.  I have no doubt that computer chess can stimulate intelligence, and that computerized chess makes the game more available, especially to those physically isolated in the past from other chess players, thus it is no surprise prodigies are getting younger.  But that shows nothing at all about computer games in general, and I can testify from watching my own children that even the so-called “educational” games, like Kerbal Space Program, are mostly an addictive waste of time destined, if not rigidly limited by parents, to turn children into pale, flabby, anti-social wrecks with much less knowledge than a child educated in a 19th Century one-room schoolhouse.

Similarly, Cowen claims that kids learn iPads by trial and error, which is true enough, and then says that “this kind of machine-based learning is driven by a hunger for knowledge,” which is utterly false—it is rather driven by a desire to play the addictive games and watch the addictive passive videos available on the iPad.  A vanishingly small percentage of children voluntarily choose to use electronic devices for actual learning.  And since the man-machine economic integration that Cowen forecasts won’t happen, to the extent children are allowed to play Dungeons & Dragons games all day to prepare for it, they will end up both uneducated and stupid.  Enhancing general intelligence may be the way to the top of today’s new hierarchy, and maybe education can help with that, but games don’t accomplish that goal, and never will.

In any case, in the real world, whether or not technology further accelerates the process, Cowen apparently believes that the non-clustered lumpenproletariat of flyover country must be brought to understand that their moral betters in the cities owe them nothing and they have no claim on them.  This is very shortsighted as a social matter.  Cowen correctly identifies that the threat of Occupy-type revolution is overstated; he rather predicts ever more economic and social division, with the lower-paid accepting inferior lifestyles in all ways, from housing to medical care, and adjusting, in part by having lots of cheap entertainment available (the real value of the computer games Cowen elsewhere praises as educational tools).  Bread and circuses—that works fine, until it doesn’t, for it hollows out a society from within.  Cowen is homo economicus—he declines to view the world in other but purely economic terms, thus creating a fatal tunnel vision.  He does not see that people do not view themselves merely as economic units; they seek transcendence and meaning, often at the cost of their economic good.  Naturally enough, he does not address the real threat—of a Trump-type revolution, but with guns, which is the more likely result of the system Cowen describes than stupid and poor, but sedated and happy people, living in tiny unheated houses in the Nevada desert.


Elon Musk (Walter Isaacson)

Tucker (Chadwick Moore)

On Marriage

On Manual Work for Men

Natal Conference 2023