Sunday, June 13, 2021

The Technosphere chokes on a chip

The technosphere, which I defined in my 2016 book Shrinking the Technosphere as a nonhuman global emergent intelligence driven by an abstract teleology of total control, has seen its interests greatly advanced in the course of the 2020-21 coronavirus pandemic, with large parts of human populations forced to submit to control measures that made a mockery of their vaunted human rights and democratic values. This is as expected: the technosphere's most potent technologies are its killing technologies, and the way it goes about using them reflects its profound hatred for all living things, especially the willful and hard to control ones. But then the technosphere started to shrink—in certain locales. It is still going strong in others, but it not to early to imagine (dare I say, predict?) how it might continue shrink and what the consequences are going to be.

In my book, I described the reasons why and the methods how we should avoid becoming trapped under the inert hulk of the technosphere. I even provided a worksheet which readers could use to track their progress in freeing themselves from the technosphere's clutches. This was, as was to be expected, to no avail. The only how-to books in this world are cookbooks; the rest are read mainly for entertainment—first alone and, later, at cocktail parties. And the purpose of writing them is to make a bit of extra money to pay baby-sitters (at least it was in my case at the time).

To understand what seems likely to unfold, have to first delve into the technosphere's ontology: what does its emergent intelligence software it run on? It turns out that, seen as a network operating system, it runs partially on human brains but mostly on various microchips, with a wide assortment of optical, electromagnetic and mechanical sensors attached. Although humans still (think that) they exercise a modicum of control over the technosphere, it is the technosphere's natural tendency to take control away from humans even unto life-and-death decisions, as evidenced by a recent event in Libya where an unmanned military aircraft autonomously made the decision to kill someone. And exercising control requires control circuitry.

Having had successful careers as an electronics engineer and then as a software engineer, I am something of a walking, talking museum of automation technology, and can take you on a brief tour of its development. The dumbest control element is the light switch. It has no memory and it decides nothing. The next slightly less stupid control element is a toggle: it remembers whether the light is on or off and when pushed turns it off or on, respectively. This is already surprisingly far along: to build a computer, we need just a few more elements. We need a threshold switch with two buttons, which, depending on what you want, turns the light on when either button is pushed (called an "or gate") or when both buttons are pushed (called an "and gate"). We also need a "not": something that turns the light off when actuated. Finally, we need an actuator; instead of turning on a light bulb, all of these elements should be able to push each others' buttons. And now we are off to the races!

All of the above can be implemented out of any number of mechanical components: mechanical, pneumatic, hydraulic, but none of these were particularly practical for automating control functions. The advent of electric circuits made possible the use of electromechanical components, enabling the great breakthrough that was Strowger switch, patented in 1891. It replaced the human telephone switchboard operator: instead of turning a crank and saying "Number 17, please!" one simply turned the rotary dial, first to 1, then to 2, resulting in a click, a pause, and then 7 rapid-fire clicks (two-digit phone numbers were the limit at the time, limiting a telephone exchange to 99 subscribers).

This system went on for a surprisingly long time. In the mid-1970s I found myself in a hotel room in Italy that was equipped with a rotary-dial phone that had a dainty little padlock on the dial to prevent guests from dialing out. But I needed to make a phone call to Russia, so I tapped out the entire long-distance number on the hook. After all, the rotary dial just actuates an interrupt switch wired in sequence with the hook.

The evolution of control circuits went from electromechanical (based on solenoids and relays) to vacuum tube-based (consisting of vacuum tube switches and ferrite cores for forming memory cells) to discrete transistor-based, to early integrated circuits (a few hundred to a few thousand transistors on a chip) and eventually to modern large-scale integrated circuits, with a recent record set by Samsung's 1 terabyte eUFS (3D-stacked) V-NAND flash memory chip, with 2 trillion floating-gate MOSFETs (4 bits per transistor). Don't worry if you don't understand what this means; just remember that it's damned impressive—because it is. But therein lies the danger. The race to build more and more powerful chips may be heading toward a cliff.

At this point just about anything—cars, washing machines, water heaters, internet routers...—has control circuits on it, all of them based on microchips. In turn, these microchips are made in gigantic factories costing several billion dollars to build. Because economies of scale are only achievable by concentrating production, each microchip is generally made at just one factory. To maintain a competitive advantage, microchips are not interchangeable. In turn, every device design that includes microchips (which by now most of them do) can only be built if every single microchip it uses is available. If that is not the case, then what is required is a very expensive redesign process to replace that one chip with another one. Often this is not economically feasible, meaning that production lines are simply shut down until all of the needed components become available.

We have already had warnings. A tsunami in Japan in 2011 drove up prices for certain computer memory chips, more than half of which were produced in Japan. A flood in Thailand caused a shortage of voltage regulators, halting car production lines around the world. And now, after a year of coronavirus emergency, there is a dire shortage of chips because of shutdowns at semiconductor factories around the world. So far Covid-19 has killed 3.75 million people worldwide, which is around 0.047% of the world's population, adding less than 5% to the normal 0.7% annual death rate. Now that multiple vaccines are available (Russia's Sputnik-V alone has been approved for use in over 65 countries) and protocols in place around the world for rapidly detecting limiting the spread of any new contagions, a reprise seems unlikely.

What does seem likely (and is already observable in many places around the world) is severe economic dislocation. Coronavirus-motivated shutdowns have caused supply chain disruptions around the world, specifically in the semiconductor industry, causing many production lines to be idled. And then come the knock-on effects. Stoppages on car production lines caused new car prices to increase. In turn, this forced rental car companies to charge more. In turn, this caused many tourists to reconsider their travel plans, causing rental car revenues to plummet, causing them to buy fewer new cars when production resumes, making it more difficult for automakers to recoup their losses.

The once expected V-shaped post-coronavirus recovery has failed to materialize; instead, what we are seeing is the onset of hyperinflation. For the very highly indebted governments, mostly in the West but also elsewhere, the standard remedy of fighting inflation by cutting spending while raising interest rates is no longer available because even a slight increase in interest rates will render them unable to pay the interest on their debt except by printing even more money, further driving up inflation.

But such knock-on effects are economic and financial; the worst ones will be physical, and will manifest themselves in the inability to maintain various life support systems that control the delivery of water, electricity, fuel, food, drugs and other essentials. Over the past decades systems that were previously operated based on paper schedules and manual operations (turning valves and flipping knife switches) have become automated, making them more efficient (in a limited sense) but much more fragile.

The electronic control systems are a layer cake of technologies. At its base are servers sitting in racks inside data centers and client systems with display screens and keyboards in the control rooms. On top of that hardware run operating systems. On top of operating systems run integrated development environments used to develop process automation tools. Finally, the process automation tools allow system integrators to configure control systems by graphically dragging-and-dropping and linking system components such as actuators and sensors and define rules and configuration parameters for their operation. Knock out any bit of any layer and the entire fragile, precarious Rube Goldberg stops functioning. The inability to replace any of these components when it fails with a compatible unit—be it a single sensor, a router or a server, forces at least a part of the entire system to shut down. And if that replacement cannot be found, then it remains down.

When looking around for a first casualty of collapse, the global semiconductor industry makes a strong candidate. It is very energy-hungry and extremely capital-intensive. It relies on a steady, reliable energy supply—wind and solar won't cut it because of their intermittency. It relies on the availability of highest-purity crystalline silicon and rare earth elements that are sourced from just a few places in the world, the main one being China. And it requires a highly disciplined and skilled workforce. The largest exporter of integrated circuits by far is China (Hong Kong and Taiwan included) followed by South Korea, Singapore and Malaysia. The US is only the first in a long list of minor players in niche markets.

It seems natural to expect that, as the market conditions affecting the semiconductor industry continue to deteriorate while the demand for critical components needed to maintain vital infrastructure systems around the world continues unabated, China will be able to exert a disproportionate influence on the availability of these components. It is quite foreseeable that the Chinese Communist Party will see the semiconductor industry as strategically important and nationalize key parts of it, fashioning it into a tool of foreign policy. The United States will, of course, pretend to be doing something about this state of affairs, making for a noisy international environment, but will not be able to prevent access to semiconductor products from becoming rationed, with China in almost complete control of the arrangements.

These arrangements are likely to be enforced by China and Russia working in tandem. China is insular by nature and can in general either trade with other cultures or absorb them. The one exception in Russia, to which China now clings like a needy girlfriend. The symbiosis is a natural one: unlike China, Russia is the opposite of insular and can digest and appropriate entire foreign civilizations. This century they are Mongols; next, Germans; then the entire Russian imperial court starts speaking French; and now English is fashionable.

As Putin famously put it, "The borders of the Russian Federation do not end anywhere." Unlike China, whose military is huge but untested in battle and uninterested in power projection, the Russians are a warrior culture that prides itself on its invincibility and that has made coercion to peace its specialty. Russia excels at building and operating huge energy, transportation and materials production systems which China needs and has the vast natural resources to continue operating them for centuries. Its fossil fuels will hold up for another half a century; after that, if all goes according to plan, it will switch to burning depleted uranium using its closed nuclear cycle technology, and there are a few thousand years' worth of it already stockpiled.

Faced with such major difficulties, the technosphere has not given up. Without filing a change of address form, it has quietly relocated and is now busy telecommuting between Moscow and Beijing. Those frisky boys at Davos and their James Bond villain wannabe Klaus Schwab are yet to get used to this turn of events. Putin and Xi have pretty much said this to their faces at their last virtual confab, but I don't think that the news has quite sunk in with them yet; let's give it time. The Germans seem to be quicker on the uptake than the rest, having understood that without Russian natural gas they would be nothing. The Americans seem to be the slowest; at this rate, it may take forever for the penny to drop. They may go down into the gurgling void all the while exclaiming that their Atlantis is not sinking!

Please buy my latest book, The Arctic Fox Cometh.