Saturday, January 25, 2014

Book Review: The Second Machine Age

The Second Machine Age, by Erik Brynjolfsson and Andrew McAfee (Erik & Andy, as they call themselves) is a sequel to the equally entertaining, self-published e-book, Race Against the Machine. The more recent book is a longer, more detailed account of the same thesis.

The argument begins with Moore's Law, which posits a doubling of computing power every 18 months. If one chooses the date that the US Government defined the economic category information technology as Year 1, then the year 2006 brings us to 32 doublings. That means computers (along with software, data communication, etc.) were 4.3 billion times more powerful than they were in 1958.

Pocket change, claim Erik & Andy. Despite the dramatic increase in computation over 48 years, computers were still something of a sideshow. Well into the aughts one could claim that the only source of economic growth was in information technology--the rest of the world still plugged along as it always had.

But now things are different. Now we have arrived at the second half of the chess board (from the story where a king promises his faithful servant grains of wheat, the amount to double on every square). For most of the first half of the chessboard, computational gains are modest if not nugatory. But as one goes from the 32nd doubling to the 64th doubling, computer power begins to take over the entire economy. Erik & Andy predict massive changes in how we will live.

So I agree with Erik & Andy. Indeed, I wrote a post last Summer (Getting Richer While Feeling Poorer) which actually serves as a pretty good description of the first half of their book. The claim (in both my post and their book) is that computation will lead to rapid automation, which will dramatically change the nature of the labor force. We both argue that traditional economic statistics, such as GDP, are no longer very useful for describing the new normal. Erik & Andy, however, include a lot more data and information in their much longer text.

Erik & Andy's thesis can be distinguished from Tyler Cowen's The Great Stagnation. Mr. Cowen argues that the world has run out of innovation--that we've consumed all the benefits of electrification, mass education, modern manufacturing, etc. Thus growth rates will inevitably decline, at least until the next big thing comes along. Erik & Andy (& me) argue that the next big thing is already here in the form of mass automation. For us, boom times are just ahead, if not here already.

Mr. Cowen recently published another book, Average is Over, in which he implicitly repudiates the stagnation thesis. He argues that indeed, computers will automate lots of jobs, but only the cognitive elite will benefit. The rest of us shlubs (80%) will be under- or unemployed and have a net lower standard of living. (His claim reduces to the argument that massive economic growth will lead to widespread poverty--a ridiculous conclusion.)

Eric & Andy specifically address the Average is Over thesis comparing Bounty and Spread. Bounty refers to the huge amount of wealth created by computer power. Spread refers to the distance between rich and poor--the greater the spread, the more the inequality. Will the bounty be distributed broadly enough to minimize the spread? They offer no definite answer, but my reading leads me to be optimistic.

It is not just the so-called cognitive elite that will benefit from computers. Indeed, a lot of doctors and college professors are going to be put out of a job, while home health care aides will have thriving careers. Erik & Andy borrow a 2 x 2 matrix to describe the result, with routine and non-routine jobs in the columns, and manual and cognitive jobs along the rows. They conclude that routine jobs, whether manual or cognitive, will tend to disappear. Assembly line work in factories (very routine) is already mostly eliminated. Increasingly, routine cognitive work (including entry-level lawyer jobs) is going the same way.

I agree with that, but I'd add some other ideas. For example, any job that requires a prodigious memory is likely to be (partially) computerized. Thus medical jobs (especially diagnosis) will rely heavily on computers and will displace a lot of humans. What is happening to lawyers today will be happening with doctors in the near future, and for much the same reason. I've posted a longer piece about the impact of automation on STEM careers here. The STEM careers that show the most promise are the skilled trades, i.e., the equipment and instrument repairmen. To use an example from Erik & Andy, the guy who can fix Baxter will make a good living.

I disagree most with Erik & Andy on their policy prescriptions. A college education, while excellent preparation for work in the 20th Century, is less well suited for the 21st Century. In particular, I think graduate work is a waste of time for all but the very top students--graduate enrollments in all disciplines should shrink by 90%. I think the baccalaureate program is too long in most cases--a two or three year college career is long enough. I think general education--such as a Great Books program--no longer makes much sense. That is something that can be pursued by the adult learner on-line as part of life-long learning.

Finally, I think math is useless for almost everybody. I took a poll in my gen ed science class (mostly art and business majors) asking how many knew how to do long division. Every hand went up. Now this is the most useless skill imaginable, yet we're still teaching it in school. Likewise, factoring quadratic equations, graphing non-linear inequalities, studying trigonometry--none of this is useful for more than 1% of the working population. Computers can do math better than you can--why are we still putting everybody through this?

Conversely, art, music, writing, performance, public speaking, cooking, dance, counseling--computers can't do any of that. Those are the subjects that should be taught in school. Save the math and science for those students who are interested in it for its own sake. There will be very few jobs in those fields.

Otherwise I really like Erik & Andy's book. Highly recommended.

Further Reading:

No comments:

Post a Comment