First-Person Judgment

??, July 1, 2024

Sometimes I’m a pessimist, on occasion I’m an optimist. But most of the time I’m a realist. Aren’t you? Is there any other way to be?

Last week’s newsletter, apart from the date controversy, presented robotics professor Rodney Brooks and his take on technology. Brooks has written and spoken a lot about how prone we are as a society to getting swept up in excessive optimism followed by excessive pessimism when it comes to tech. He worries we’re going too far in the direction of pessimism about AI, and will overcorrect with excess pessimism. The initial excitement about AI and its adoption rates have been exceptionally high, and he suspects they imply a big correction. His take looks to me like that of a realist.

Brooks sees this extreme hype cycle to be a major challenge to further development, as the overly optimistic response has inspired over-investment. As a response, too much money will be withdrawn when the losses are booked. This is the familiar pattern best known from the early internet Dotcom Boom and Bust of 1999/2000.

For my part, I believe most of our environmental policies right now are based on an excessively optimistic view of what tech and engineering can do. For instance, when it comes to battery technology, there’s an unfounded assumption that Moore’s Law will produce exponential rates of improvement. But not even computer technology can enjoy Moore’s Law forever.

Gordon Moore, cofounder of Intel, posited back in 1965 that the number of transistors put on a computer chip would double every two years, with a commensurate increase in computer performance. Such a rate of improvement was and is astounding—it is exponential. Yet it can’t go on forever. Some say it no longer even applies in computer chips, since the physical limits have been reached, as circuits are now only atoms wide. No one has even the wildest inkling of how to manufacture switches smaller than atoms, but that’s what Moore’s Law would require if it still applied.

The Moore’s Law improvement rates have never been posited for other areas of technology. For instance, we don’t expect the speed of transportation to increase exponentially. Our species achieved a tremendous improvement in speed when the horse was domesticated some thousands of years ago. Horses allowed us to do more work, and they allowed us to travel faster than our own legs could carry us for short stretches, but there was no assumption this was the beginning of exponential speed increases.

The next major gains to speed came with the advent of steam machinery and railroads in the 19th Century. These improvements eventually spilled over into rail-free motor vehicles and then aircraft, culminating in jets and manned rockets, speeding us along at multiples of the speed of sound. If we had assumed a Moore’s Law’s rate of improvement, we would already have achieved the speed of light and beyond, well past the realm of speculative science fiction. Achieving such speed increases would call for technological advancements in materials and physics that we have no discernible means of attaining.

Nevertheless, there is an implicit belief in contemporary society that energy stored as battery power (per unit of weight) will improve at an exponential rate. This sounds entirely fanciful, not least to chemical engineers. And battery storage of electricity is foremost a set of engineering problems to solve.

The normal rate of improvement measures progress in ever smaller increments. It’s the classic economic concept of diminishing returns. There still is improvement, but the early gains are the easiest. Subsequent steps cost more in terms of money, materials, and time spent on them. Internal combustion engines (ICEs), by way of example, have continued to become more efficient as newer technologies have come along. Today’s cars have a fuel efficiency that is worlds apart from that of Henry Ford’s Model A. Yet the improvements have never been exponential, while the biggest gains were made early on. Early ICEs were small and weak, but consumed a lot of fuel for every watt or horsepower of work done. The output per fuel burnt improved a lot in the early decades, as engine designs improved quickly, not to mention the improvements in petroleum refining and in metals for engine parts.

The materials used in building cars today are also worlds apart from those used in the early cars. Engine blocks are made of alloys of metals at levels of purity that were technologically impossible a century ago. Other vehicle parts are made of materials that hadn’t even been dreamed of in the early 20th Century. And all of them are necessary for modern vehicles’ energy efficiency.

Today’s battery technology has also come a long way from that of over a century ago. The lead-acid-based batteries found in cars is very old technology, and it is still in wide use. Battery tech improvements since the 19th Century have not only fallen short of Moore’s Law, they have not fixed all the problems that ancient technology solved: the ability to store and access electricity instantaneously over a very wide temperature range. Plus, the chemicals for lead-acid batteries are relatively cheap still. Yet those component costs are going to increase rapidly as more and more battery power is mandated by inflexible (and unrealistic) government standards. In other words, there are limitations on producing things at a global scale.

The improvements that led to nickel-cadmium and then to today’s lithium-ion batteries required changes in global mineral mining, extraction, and manufacture before they were cost effective and became widespread. Just in terms of raw materials, every chemical element has its unique geographic distribution, its own availability profile. Any new battery technology calls for nothing short of a global industrial revolution before it can become significant. That is a major hurdle to quick rollout.

In the computer realm, improvements to manufacturing silicon chips were a large part of the improvements, but silicon remained a major ingredient. For performance, there was no central authority imposing new benchmarks on the industry from the outside, but rather industry-leading corporations with access to capital who could wring more out of existing material mixes. If the chip business had needed to rely on discovering new ways to use different chemical inputs for each leap in computer power, Moore’s Law would never have been possible.

Today, we still tend to think exponential improvements can continue in computing. That remains to be seen. But no one expects exponential—or for that matter, linear—improvements to the speed of transportation. Not even the government planners who appear to believe you can mandate energy production into existence on a national scale without significant cost. Otherwise, they would at least be mandating a return to supersonic passenger air travel.

As is the case for every industry apart from computers, advancements in battery technology are highly unlikely to become exponential, it is safe to say. That would be the realistic view.