Erik BrynjolfssonRecensioner

The more I read, the less I know how to prepare my children and students for this rapidly changing world.

All in all, disappointing. Common sense opinions about the role of technology for the future of work and not much else.

Analysis of the impact of machines on economy, since the industrial revolution. Nothing really new under, some useful recaps and historical references.

Although the examples are a bit out-of-date, the authors do capture many of the relevant trends shaping the evolution of the workplace. The writing is incredibly approachable while incorporating an array of studies from economics, sociology, industrial relations and beyond. Their section on the superstar economy was particularly compelling.
Because they discussed that barriers (both psychological and financial) to organizational process change inhibit the productivity gains from new technologies, I can't help but wonder whether the current global pandemic and economic devastation will serve as an exogenous shock to catalyze many of the changes they discuss.

With the proper amounts of bombast and self-promotion, prophecy can be an extremely lucrative field. A lot of books about the future of American economic growth fall into either Stagnation or Singularity camps, trying to show that America's future potential is dismal due to the misguided economic philosophies of the ideological villains of your choice (see the New York Times or the Wall Street Journal's op-ed pages) or simple exhaustion of easy ways to generate sustained growth (see Robert Gordon or Tyler Cowen), or is actually spectacular due to the magical properties of game-changing innovation X, Y, or Z (see Ray Kurzweil). In contrast, Brynjolfsson and McAfee's earlier book Race Against the Machine was a sober, data-driven overview of what they thought the likely effects of increased automation on the labor force would be, with interesting case studies and plenty of good data. Its major flaw, in my view, was an ending solutions chapter that spent more time on perennial Silicon Valley wishlist items like reforming the patent system or allowing for more H1-B visas than on engaging with the political process. While those wishlist items haven't gone away, this sequel not only offers a greatly expanded take on their earlier analysis of technological progress, but a broader and more carefully thought-out list of possible solutions to problems that automation will cause for many workers even as the economy as a whole benefits greatly.

The first section of the book is devoted to the three characteristics of modern technological progress: exponential growth, large amounts of digitized information, and constant remixing of old ideas into new ones. From Google's self-driving cars, to Apple's Siri voice recognition system, to IBM's GeoFluent translation software, to IBM's Jeopardy!-beater Watson, and more, it's obvious that while artificial intelligence might have had limited progress for a long time, it's come very far, and further progress seems like it will come much more rapidly. This is the "second half of the chessboard" metaphor from their first book, taken from an Indian story about the inventor of chess, who asked for a seemingly-simple reward for his game from the emperor: one grain of rice on the first square of a chessboard, two on the second, four on the third, eight on the fourth, and so on. The emperor agreed, not realizing that while the amount of rice he'd have to hand out would be fairly manageable for the first half of the chessboard, by the second half he would be in real trouble. A classic example of this is Moore's Law, but in addition to semiconductor density, other measures like energy efficiency, hard drive cost per MB, and supercomputer speed also follow exponential growth patterns.

This allows for truly immense amounts of data to be processed, which enables all sorts of useful stuff that wasn't possible before. An example is Waze, a now Google-owned company which, thanks to large amounts of real-time input from its rapidly growing userbase, shows you the best route around traffic and makes traffic forecasts obsolete. The combination of network effects, lots of data, and the near-zero cost of reproducing or transferring that data, not only ensures that data scientists/analysts looking for hidden insights will have plenty to do, it also means there are plenty of opportunities to go back to old ideas and do something new with them. While plenty of modern innovations seem easy to scoff at - with Waze, for example, surely an app that speeds up my commute by 5 minutes can't really compare to the invention of the internal combustion engine - scientific fields are expanding so quickly that there are plenty of opportunities for many people to make small improvements that eventually add up to large rises in our standard of living.

The second section of the book explores bounty, their word for the value of goods and services this new Revolution has given us, and spread, their term for the distribution of that bounty. For a long time there's been a controversy over the seemingly-negligible part the IT revolution has played in the official measurement of official productivity and other economic statistics - individuals and firms have spent a lot of money on hardware and software, yet that investment, as measured by GDP growth or other measures of output growth per unit of input, has been unimpressive. Brynjolfsson and McAfee are firmly in the camp that holds that technology has improved life, it's just that official measures don't capture that very accurately. They compare the adoption rates of electricity to IT and find interesting similarities in how long it's taken for each to start showing up in the numbers, and also quote the famous Robert F Kennedy line about GDP, that it "measures everything, in short, except that which makes life worthwhile." They list four types of intangibles that won't show up in national accounting statistics: intellectual property, organizational capital, user-generated content, and human capital, and note that all of these have been exploding recently. To take one example, Wikipedia has been a death sentence for traditional encyclopedia makers, yet it's a non-profit website. In conventional GDP statistics, it has destroyed millions, perhaps billions of dollars, and yet it's been an enormous boon to everyone who wants to look something up quickly. Email has been bad for the post office. Skype is bad for phone companies. Music publishers hated Napster. And so forth.

All this consumer surplus is great, but it has consequences for where profits flow. We all enjoy bounty to some degree, but it's not spread very evenly. They use an example from the field of photography:

"While digitization has obviously increased the quantity and convenience of photography, it has also profoundly changed the economics of photography production and distribution. A team of just fifteen people at Instagram created a simple app that over 130 million customers use to share some sixteen billion photos (and counting). Within fifteen months of its founding, the company was sold for over $1 billion to Facebook. In turn, Facebook itself reached one billion users in 2012. It had about 4,600 employees including barely 1,000 engineers.
Contrast these figures with pre-digital behemoth Kodak, which also helped customers share billions of photos. Kodak employed 145,300 people at one point, one-third of them in Rochester, New York, while indirectly employing thousands more via the extensive supply chain and retail distribution channels required by companies in the first machine age. Kodak made its founder, George Eastman, a rich man, but it also provided middle-class jobs for generations of people and created a substantial share of the wealth created in the city of Rochester after company’s founding in 1880. But 132 years later, a few months before Instagram was sold to Facebook, Kodak filed for bankruptcy."

In other words, rather than benefit large numbers of people in many communities through plentiful jobs, profits are flowing increasingly to a few people in a few places like Silicon Valley. Anyone who paid any attention to the 2012 Presidential election is familiar with the notion that median wages have been stagnant or declining for many groups for quite a while, and net wealth has taken a shocking decrease since the Recession in particular:

"Between 1983 and 2009, Americans became vastly wealthier overall as the total value of their assets increased. However, as noted by economists Ed Wolff and Sylvia Allegretto, the bottom 80 percent of the income distribution actually saw a net decrease in their wealth. Taken as a group, the top 20 percent got not 100 percent of the increase, but more than 100 percent. Their gains included not only the trillions of dollars of wealth newly created in the economy but also some additional wealth that was shifted in their direction from the bottom 80 percent. The distribution was also highly skewed even among relatively wealthy people. The top 5 percent got 80 percent of the nation's wealth increase; the top 1 percent got over half of that, and so on for ever-finer subdivisions of the wealth distribution. In an oft-cited example, by 2010 the six heirs of Sam Walton’s fortune, earned when he created Walmart, had more net wealth than the bottom 40 percent of the income distribution in America. In part, this reflects the fact that thirteen million families had a negative net worth."

In English, that means all those Occupy Wall Street slogans about the 99% are more correct than their detractors would like to believe, but the overall picture is (slightly) more complex than "the 1% are stealing everything". Even beyond the decoupling of wages and productivity, there's been a global fall in the labor share of GDP, meaning that wealth is flowing more to owners of capital. In many sectors we now have a superstar economy, where there's enormous demand for a few popular things at the expense of many less-popular things. This entails a shift from returns for absolute performance (meaning that performing at 90% of the level of the best gets you 90% of the return) to returns for relative performance (meaning that no one wants the tenth-best app in whatever field, so you get nothing). This can be partially counteracted by long tail-type economies, which make it possible for relative low-performers to scrape by, but they're not very lucrative because of the power of network effects and technological lock-in. Even if Windows Phone is on paper basically just as good as iOS or Android, no one cares, and though everyone derives some benefit from the ubiquity of smartphones, profits in the industry flow to very few firms, and within those firms, even fewer people.

This shift from normal/Gaussian distributions of wealth and power to 80-20/power-law distributions is profound; Brynjolfsson and McAfee cite Acemoglu and Robinson's work in Why Nations Fail on the relationship between political institutions and economic distributions, and how exclusive political systems that are set up for the convenience of a small elite not only don't grow very fast but are also terrible places to live for the masses. I would have liked to see them address Paul Krugman's points in The Conscience of a Liberal about how political movements can drive and encourage this re-peasantization process, though I can understand their desire to avoid seeming too strongly partisan or ideological. Tyler Cowen's recent Average Is Over was a good example of how to be too ideological in the wrong direction about these same concerns, arguing that this return to the Gilded Age will be much more pleasant than the original Gilded Age. Sure, large numbers of people will be permanently excluded from labor markets and won't be able to meaningfully participate in the political process, but all the cool new technology means that that won't be so bad, or at least not bad enough to get too upset about.

Unfortunately, Brynjolfsson and McAfee's take on the question of what happens when automation starts to put significant numbers of people permanently out of work is more pessimistic. Even though people have scoffed at this vision of technological unemployment for hundreds of years (see the history of the Luddites and the "lump of labor" fallacy), this time could be different. There are three possible mechanisms for destroying jobs in this way: inelastic demand for goods (this could actually be good, as people would be able to voluntarily choose to work less while still producing the same amount of output), too-rapid change (it might simply take too long to re-skill certain types of workers), and severe skill inequality (some people will just never be able to produce value greater than what a machine could do). This new permanent underclass will be subject to the sorts of social pathologies that got people transported to Australia in past eras, but options in the future will obviously be somewhat more limited. Ironically, the kinds of jobs easiest to automate are also the kinds easiest to offshore, so America might get a breather from offshoring and be able to watch and learn from what happens as large-scale automation in companies like Foxconn gets field-tested overseas before coming here. Of course, Freestyle chess, where humans and computers collaborate to accomplish goals, could be a model for the economy as a whole, but it seems more likely that many people will simply be automated out of a job and left to their own devices.

In the third section they have two groups of solutions, of which the first recapitulates much from Race Against the Machine. In the short-term:
1. Teach the children well
- Use MOOCs, which are both cheaper and provide more opportunities for data-driven feedback
- Raise teacher salaries in exchange for more accountability, coupled with longer school days and a longer school year
2. Restart startups
- Startups provide most new net jobs, but the rate of new startup formation is dropping quickly. "Regulations" might be to blame
3. Make more matches
- Do a better job of matching workers to prospective jobs to reduce frictional unemployment as much as possible
4. Support our scientists
- Reform intellectual property laws by lessening absurdly long copyright terms
- Offer more prizes for research goals, to bring in people who don't fit into the regular grant process mold
5. Upgrade infrastructure/human capital
- There's lots of externalities to improving our terrible infrastructure, even beyond arguments about Keynesian stimulus
- Welcome more high-skill immigrants who are currently going to other countries, and also reform the byzantine/broken immigration process
6. Tax wisely
- More Pigovian taxes that tax bad things, like congestion or pollution taxes
- Consider a land tax or a VAT to fund social programs instead of relying on labor taxes
- More taxes on being a superstar, like higher tax brackets

The second group is new, and to my mind more adequate to the scale of the issues raised previously in the book. In the long-term:
- Build on capitalism and unlimited technological progress without abandoning or attempting to fundamentally restrain either
- Consider a Universal Basic Income, to prevent Voltaire's social ills of "boredom, vice, and want"
- Alternatively, consider a Negative Income Tax like a greatly expanded EITC to encourage work
- Find better ways to use the strengths of humans and machines together, as in Amazon's Mechanical Turk
- Bring marginal people into the labor force via the peer economy, e.g. TaskRabbit, Airbnb, Lyft
- Encourage new ideas (a national mutual fund, designate some jobs "human-only", institute "made by people" labeling similar to that for organic foods, use massive federal hiring a la the Civilian Conservation Corps)

Unfortunately they don't offer much in the way of suggestions on how to move these ideas though Washington. Hey, they're nerds, not lobbyists! Well, any book that attempts to grapple with the consequences of something as world-changing as artificial intelligence on a large scale should certainly be able to offer some pointers on how to get the Republican Party to start offering real solutions to problems that don't boil down to tax cuts for the rich. This kind of naivete is unsurprising, yet still disappointing coming from such smart guys. The additional analysis in the first two sections and the broader range of solutions in the third means that this is a much more complete and useful book than its predecessor was, and while I certainly wouldn't say that this is the final word on the possibilities and pitfalls of large-scale automation, it's as good a starting point as you're likely to find for a while.

I love short, data- and graph-heavy books like these, that tackle important subjects in greater depth than the blog format allows but at greater brevity than a 400 page tome. Even when I have big problems with them, like I do with this one, they're usually very brisk yet well-written and thought-provoking. The book's central subject is an important one - the effects of automation on the economy of the future (mostly the US, but the results are probably more broadly applicable). It's broken up into 4 quick chapters, of which the first 3 are descriptive and the final one proscriptive. I'll bullet-point the logic of the first 3 chapters:
- Unemployment is extremely high by historical standards and doesn't show signs of falling swiftly
- This could be due to three broad causes: mismanagement of aggregate demand (e.g. Paul Krugman's view), a general stagnation in technological progress (cf. Tyler Cowen), or a fundamental shift in the kind of jobs available
- While not exactly ruling out the first two causes (though Cowen believes that modern technological progress is nowhere near as transformative as the steam, electricity, etc that we got during the Industrial Revolution), the authors think that progress in automation deserves a closer look
- We're getting into "the second half of the chessboard", where accelerating technological progress in automation can be seen in things like IBM's Watson or Google's self-driving cars
- This progress will make an increasing percentage of traditional jobs either obsolete or so low-paying as to make no difference
- It's not obvious that new jobs which will be created will be either as numerous or as remunerative, especially since technological progress is sort of "outside" the business cycle
- Even then, those new jobs which are remunerative will require lots of skills that will be hard for much of the population to acquire
- Even worse, we live in a superstar/plutocrat economy where most gains flow to a small number of people, leaving declining real living standards for everyone else
- This means that the future for most people looks grim, in fact so grim that the authors quote Gregory Clark thus:

"There was a type of employee at the beginning of the Industrial Revolution whose job and livelihood largely vanished in the early twentieth century. This was the horse. The population of working horses actually peaked in England long after the Industrial Revolution, in 1901, when 3.25 million were at work. Though they had been replaced by rail for long-distance haulage and by steam engines for driving machinery, they still plowed fields, hauled wagons and carriages short distances, pulled boats on the canals, toiled in the pits, and carried armies into battle. But the arrival of the internal combustion engine in the late nineteenth century rapidly displaced these workers, so that by 1924 there were fewer than two million. There was always a wage at which all these horses could have remained employed. But that wage was so low that it did not pay for their feed."

Yikes. So, what is to be done? This last chapter is a predictably mixed bag. The note that in theory, long tail-type economies of scale give small producers an edge, as does the app economy and things like home 3D printing, though I'm not sure if it's feasible to have an economy of people making new Twitter apps and Etsy products for each other. For education, they recommend paying teachers more (good), instituting dubious reform measures like eliminating tenure (???), keeping kids in school for longer (probably good), importing more skilled immigrants (good), and embracing MOOCs (possibly good). For entrepreneurship, make it easier to start businesses (good) and streamline burdensome regulations (sounds great, but these sorts of bullet points are always too vague). For investment, they like more infrastructure spending (good) and R&D spending (good). For laws, they like extremely "flexible" hiring/firing (hmm...), cutting employer payroll taxes (maybe good, but what happens to social programs funded by those taxes?), decoupling employment from benefits like health insurance (good), not regulating new network businesses (unlikely), ending the home mortgage subsidy (good), ending too big to fail (good), fixing the patent system to end patent trolling (good), and shortening copyright periods (good).

Though many of these suggestions are good, this seems like the kind of techno-nerd perspective that plays well to other thinkers but doesn't engage much with the people who are actually being downsized and automated. It's possible to think that the gains from technological progress will tend to only accrue to a few, but that this is really a policy choice driven by plain old class politics rather than something inherent to the iEconomy. Mark Zuckerberg can get really rich off of Facebook, but the means by which he did so (essentially private ownership of a large corporation via control of most voting shares) are centuries old. I'm willing to grant that their "third way" thesis that the current age of technological progress might be driving part of unemployment, but even though I don't think the solution to the challenges of modern technological progress should be exactly the same as New Deal policies to spread wealth and opportunity after the Depression, one looks in vain in this book for alternative suggestions that would help give workers more control over their own destiny. It's definitely a good thing that smartphones are becoming more powerful and more widely available, but it's not really a mystery why unemployment is going up if the only means by which people can react to modern changes in their lives is the same limited means they had during the Gilded Age. Man might be, as they quote from a NASA report, "the lowest-cost, 150-pound, nonlinear, all-purpose computer system which can be mass-produced by unskilled labor," but we can still vote, right?

Perhaps their upcoming book will expand further.

There were some worthwhile ideas in here, but nothing really novel or deeply insightful, and it could have been communicated in a few pages instead of a book. If you are really interested in TED Talks and feeling important by learning a superficial amount about a new problem and some potential solution, maybe you should read this, but I'd rather actually learn something meaningful and do something with it.

If you are in the top 10% of a first world country the future is going to be awesome, like you won't believe. More of the same for everyone else, albeit with better free entertainment. My only complaint is that the book gets highly repetitive whenever the authors get over-enthused.

Concise overview of why digital technology will cause a lot of unemployment.

Ecrit par deux des meilleurs spécialistes de la transformation numérique, ce livre met à la portée de tous les évolutions que l'on constate aujourd'hui, notamment en matière d'intelligence artificielle, algorithmes de plus en plus puissants, apprentissage profond.

πολύ διαφωτιστικό, εξαιρετική μετάφραση

I Took this book for holidays reading ... and absolutely not disappointed !
Very nice insights around Machine écosystem ! Explanations and examples are easy to follow.
A must read for anyone interested in Digital world and wondering about the future of it !
Really a great book and must read ;()

July-August 2017

How the digital revolution is accelerating innovation, driving productivity, and irreversibly transforming employment and the economy

More economics than technology

How has technology affected human culture and economics? How will the rapid progress of digital technology do so in the future? These are the questions this book attempts to answer, albeit not in any great depth. Primarily it's about the economic impact of increased productivity due to automation. It highlights some likely challenges ahead (such as unemployment and the growing spread in wealth between capital owners and people who work for a living), and it offers suggestions on how those can be met. I can't say it brings up any major points I have not thought of before, and my only real disappointment in this book is that it offers nothing truly bold, insightful, or innovative. Still, it is a good read. I recommend it.

The purpose of the book is to talk about how machines are and will change our lifestyle in the coming years. With Internet of Things being on the rise, I do see things change very drastically (in a positive way) for us very soon. I was impressed by the 'second half of the chessboard' analogy. I could relate to it being in the technology field myself.

The authors hint at improving productivity - throughput per worker - in the early chapters as a means of achieving this. Higher levels of computation, digitization, and recombinant innovation are the tools that are making these breakthroughs happen.

They do place an emphasis on better learning techniques, both in school and after. A mention of MOOCs and the way students can make the most of it was definitely a revival for me as I've read it earlier in other books.

Few of things mentioned in the book that the authors think as 'wonders of the future' are actually happening right now, but that could be my timing of reading the book.

Disappointing look at the challenges we humans have competing for work against machines. It says very little that is actually new½

What a leftist hogwash!

The authors make two assertions: 1) In the long run machines will become so smart that they can handle almost every tasks which until now required human input. 2) Because of so many machines doing so many tasks prices of all things will fall dramatically, will tend even to zero!

All right so far. But suddenly a deep red Marxist jumps up and says: "The more machines doing these tasks the more profits will accrue to capitalists!"

Why this Marxist is so red? Well, he is unable to uphold this nonsense, because when all products are getting cheaper every year how should it possible for capitalists to claim even higher profits?

The authors don't bother about that very long. For them it's important to conclude further: "In this coming second machine age we need a compensation for the expected masses of unemployed!"

And this is supposedly the gist of the matter: Write some alluring lines that my fellow-socialists can believe that it is necessary to 1) print even more money and 2) use it for an unconditional basic income for all!

Hallelujah!

Avoid this in favor of The Second Machine Age.

Ok ... not much else to say.

In all, a very good and informative read, although the key ideas are not new to readers of Jeremy Rifkin's 1994 book "The End of Work", which, unfortunately, is not mentioned.

I especially liked the argument that "growth" is increasingly inadequately captured by GDP growth, and the point that the present fiscal system is too much labor-oriented. In general, the diagnosis was excellent. The solutions outlined by the authors, however, were much too short-term in my eyes. Especially since the authors stress that we are at an "inflection point" of history, focusing on quick fixes of the status quo (better education etc) is a little myopic. We need to be prepared for a largely laborless society within our lifetimes, which will require huge changes in the distribution of income, as the authors themselves acknowledge. This big transition will take a lot of time, so it must be started now. The authors were too light on outlining the long-term solutions. For example, how are governments going to finance negative income taxes for the legions on un(der)employed, and the necessary investments in science and infrastructure? I would have liked more detailed visions on the solutions for the "android experiment".

Lastly, for a book about technology, the ebook version is funny in that the final 15% consist of a (completely useless because the keywords are unlinked) index; it's also highly misleading as the main text already ends at 67% of the ebook. In general, the book makes the impression that it could have used another round of editing.

I work in IT and have spent most of my career working to stay ahead of the awesome creative destruction that pretty much defines the industry. This book rang true.

The authors set out to give the lay of the land - and do so quite well - and then compare it to previous bursts of innovation (industrial revolution, electricity). They build a compelling case for what is yet to come and back up their conclusions with mountains of data.

There's a lot of optimism about what the future of technology looks like. There's a lot of hard news for people whose job can be easily replaced by a machine. And there's a lot of advice for how we can best prepare.

If this book was written in the late 18th century, it would tell us that our career as a field laborer might not be the best option, but that now would be an excellent time to invest in steam engines and railroad stock.

If you're already in the workforce, this book will give you a valuable glimpse of what is coming. If you're in high school and are taking the important first steps of the rest of your life, drop what you're doing and buy this book right now. It may well change your life.

Brilliant but disturbing view of how our society and economy are changing in front of our eyes.

The authors wrote this book as a follow-on their previous book about technology and the economy, "Race against the Machine," which I found very thought provoking. In The Second Machine Age, they look at how things once considered un-automatable like driving and medicine may well not be and how that will affect the economy. I've read a lot about driver-less cars by Google (and others) and can't wait until I can either work or sleep in my car while it takes me where I need to go. What I had not given much thought to was how this will affect the economy. In this book, the authors explore some of the consequences if jobs like long-distance trucker are no longer ones done by humans. So too, in medicine where now there are interesting developments with IBM's Watson technology being used to diagnose patients. How will that affect doctors and the healthcare field as a whole? At one level, these are not immediate concerns, but at the same time the technology is moving very quickly and will have consequences for all of us. This book is well worth reading for anyone who likes to think about future technology and how it will affect us.½