There are several areas where semiconductor technology and society come together. One is the future of work. As computers and deep learning get more powerful, there are a lot of jobs that potentially can be handled more economically by computers. For example, in Silicon Valley we look at the fast progress that autonomous vehicles are making, and we all know people who have Teslas who already happily commute to work leaving much of the driving to their car while they look at email. But there are about 3.5 million commercial drivers in the US. I'm not quite sure what is included in this number, but it doesn't matter that much: it is a large number of people whose jobs are likely to be automated away in the next...pick your number of years. But semiconductor technology has enabled the internet, obviously, and that has already had a major impact on work in the last twenty years or so. Branko's Elephant Chart I recently came across a chart that has been around for ages, and deserves to be more widely known. Toby Nangle of Columbia Threadneedle Asset Management, called it "the most powerful chart of the last decade." It was created by a World Bank economist called Branko Milanovic and first published in a 2012 World Bank working paper . It has come to be known as "Branko's elephant chart" because of its shape with a big body and a trunk. Here is how it appears in the paper: It needs a bit of explanation. Firstly, this is worldwide data, not data from just the US or just the developed economies. Across the bottom is where in the world income distribution per individual is. The leftmost dot is the lowest 5% ($1/day type of level) and the rightmost two dots are the highest 5% and 1% (which probably includes you, as this is worldwide data). The blue line shows how the income has changed from 1988 to 2008. The very poor remain very poor, then up to about the 65th percentile people have got a lot richer, up to 80% richer, then from there up to about the 90th percentile, income has stagnated, and above that (the trunk) people have got a lot richer again. Who are these people? Of course there are individuals all over the place, but what this chart broadly shows is that a lot of jobs that used to be done by the middle class (70-90th percentile) in rich countries, is now done by the middle class (20-70th percentile) in rapidly industrializing countries, especially China, at lower wages. There is another trend that is not captured in these charts, namely increasing automation of industrial manufacture. With all the rhetoric around, it is easy to miss that US industrial output is at an all-time high. Agricultural output is at an all-time high, too. But neither activity requires that many people any more. Manufacturing employment actually peaked a long time ago, in 1979. An example closer to home is that when I started in the semiconductor industry, fabs required a lot of people since automation of the manufacturing process was low. Wafer lots were hand-carried between tools, for example. The settings on the tools would be dialed in manually. A modern fab is almost completely automated. It is not just in the US, either. Since the iPhone 6 was released, a couple of years ago, Foxconn has replaced half their workers with robots. The numbers are big, too. They reduced their iPhone workforce from 110,000 to 50,000. The Great Convergence An interesting book by Richard Baldwin called The Great Convergence has a view on exactly what is going on. You don't even need to read the entire book to get the basic ideas, since Richard has provided a 30-slide PowerPoint deck . There are three different costs that constrain globalization: The cost of moving goods (trade costs): These have fallen, starting in 1820, and speeding up with the invention of containerization and better logistics software The cost of moving ideas (communication costs): These fell rapidly with the arrival of the internet The cost of moving people (face-to-face costs): We are unlikely to fly to China for our haircuts, but medical tourism is already a thing From 1820 to 1990, there was what Richard calls the great divergence, where more and more of the share of world income was captured in the rich countries. Moving ideas was just too difficult and so innovations in the US would stay in the US. As Richard puts it, "G7 labor had a monopoly on G7 knowhow." You can see that in the green dots on the chart below. Then from 1990 has been the great convergence. The cost of communication fell so that the link between innovation and implementation was broken, the US could innovate but it could use China to manufacture. You can see that in the red circles in the chart below. These only show half the story, the share of value captured by the rich countries has fallen, but unseen is that the share in developing countries has risen. The share in US and the share in China are getting closer, hence "convergence". It is this that drives the elephant chart. The people doing the manufacturing in China have moved from subsistence agriculture to the lower end of the global middle class. The people who used to do the manufacturing in the US have struggled. And at the high end, the trunk of the elephant chart, are the people who create and manage the innovation. If you have an iPhone, look on the back: "Designed by Apple in California, Assembled in China." The trunk and the body of the elephant. But technology is probably the more important factor than globalization, I think. For example, US steel production is down about 20% since 1970, despite consumption of steel increasing, with the gap filled by imports. But steel employment is down 80%. Most of that 80% is due to technology, not trade. As computer learning gets more powerful, many more jobs will be subject to being replaced by machines. None of us can be sure our job won't be next. We have been through a series of transitions from agriculture, to manufacturing, to today where most of us work in what is called the service sector. The big question is what comes next? And if the answer is something along the lines of "the end of work as we know it," how will society structure itself? Previous: RISC-V: Codasip, BaySand, and More![]()
![]()
↧