Zero Genie

Trying to anticipate the future of technology is a fool’s errand, is it not? After all, even Dr. Vannevar Bush, a renowned architect of the American science establishment, predicted back in 1945 that intercontinental missiles would be impossible for many years. And Thomas Watson, the President of IBM (yes, the artificial intelligence system “Watson” was named after him), is said to have predicted in 1943 that there would be a world market for maybe five computers. The list of laughably wrong predictions about future technologies is endless. Surely we should not be too hard on ourselves when ambitious weapon programs keep failing because, in part, the necessary technology turns out to be too far away. How could we know?

I disagree. The future of military technologies — with implications for the future of the warfare itself — is not an unknowable and random mystery. Granted, examples of erroneous forecasts are numerous. And yet, quantitative research shows that the accuracy of predictions in military technologies — even long-term predictions looking 30 years into the future — are correct surprisingly often, about 70-80 percent of the time. Even the numerical measures of systems’ performance grow over time in a fairly consistent, reasonably predictable manner — the figure above is but one example.

Figure 1: A composite measure of performance attributes for multiple direct fire system grows in a consistent, regular manner over the last 700 years. (Journal of Defense Modeling and Simulation)

Decision-making about military technology development programs — and their specific objectives — does not need to be a triumph of hope over experience. It can — and should — rely on persistent, disciplined, quantitative, history-based technological forecasting.

Accuracy of Predictions

But how do we know that technology forecasts can be sufficiently accurate? In research with my co-author, I explored the average accuracy of long-term forecasts about future military technologies. Here, by “long-term” I mean about 20-30 years. Why so long? After all, commercial technology forecasts tend to focus on significantly shorter time horizons, only up to 10 years into the future.

Become a Member

Unfortunately, it has become common for a major defense acquisition program to take on the order of two decades from concept development to initial operating capability. Even before that, it takes another 10 or more years to develop the necessary foundational science. That’s why a 20 year or longer horizon is often important for military technology forecasts.

In gathering the data, we happened to be lucky: It turns out that back in the 1990s, as the Soviet Union collapsed, a lot of intelligent people were gazing into the crystal ball trying to figure out what would happen with all things military. And for some reason, many of them liked to make their predictions about the year 2020. (Perhaps they liked the sound of the phrase “2020 Vision”.)

We collected a number of such published predictions. For example, that swarms of armed unmanned aerial vehicles (or loitering munitions) would be able to destroy numerous ground targets, or that some tank munitions would be laser-guided. Then we asked ten highly experienced and well-credentialed military technologists to judge whether the predictions came true. On average, the experts’ assessments showed that the predictions were 76 percent true. That’s a surprisingly high number.

We also explored a somewhat different question: Even if a particular prediction has not yet materialized, does it represent a promising direction for research and development, and has it exhibited significant progress by this time? We found that by this measure, 89 percent of forecast statements were good.

Another interesting finding was that some technology categories exhibited much higher forecast accuracy (with strong statistical significance) than others. Specifically, the average accuracy of forecasts related to “informational technologies” (i.e. technologies for cyber and electronic warfare, sensing and information collection, and command and control) was 87 percent. And forecasts of “physical technologies” (i.e. line-of-sight effects, non-line-of-sight effects, protection, and platforms) had the average accuracy of only 65 percent. These numbers are broadly similar to what other researchers have found about non-military technologies.

The Long Trajectory

To be clear, the research I describe above was about qualitative capabilities or features, not quantitative. It did not touch on forecasting numerical characteristics of future technologies. For example, a forecast like “some tank munitions will be laser-guided” can be true or not, but it does not say anything about the quantitative range of the munition, how much armor it can penetrate, and so on. How about giving us some hard numbers?

Well, it turns out that in some ways the numbers could be even easier to forecast. Many law-like quantitative regularities are known to apply to technological systems. For example, certain performance measures of technological systems often exhibit exponential (or similar) patterns of growth over time, meaning that if you take a logarithm of a measure and plot it as a function of time, the curve will be a straight line. A particularly well known example of such a regularity is the Moore’s Law. It states that a performance measure of a computer chip doubles approximately every two years. Many other technologies follow a similar law of exponential growth. Even widely different — but functionally similar — technologies end up forming a surprisingly steady, law-like trajectory of development.

Recently, I researched whether such a regularity might describe a diverse collection of mobile direct-fire systems, over a long, multi-century history. I considered widely different families of technologies that span the period from 1300 to 2015: Soldiers armed with weapons ranging from bows to assault rifles, foot artillery and horse artillery, towed anti-tank guns, self-propelled anti-tank and assault guns, and tanks.

Quantitative analysis shows that a single, uncomplicated regularity describes the historical growth of this extremely broad collection of systems. Remarkably, a fairly simple formula when applied to multiple, widely different weapon systems — from a bowman to a tank — produces numbers that all fall approximately on the same curve, a function of time. The key part of this empirical formula turns out to be the maximum kinetic energy that the system can potentially direct at the target in unit time and per unit mass of the overall system. That measure of energy took about 60 years on average to double before the 1830s, and about 15 years after.

You can see this curve in the figure at the beginning of this article. Essentially, it is a combination of two subsequent exponential laws: One straight line from 1300 to the 1830s, and another straight line between the 1830s and the current time. If the latter line holds (and so far there is no particular reason to doubt that it will), it can be used to forecast some of the characteristics of future mobile direct-fire systems. It is not going to be precise — indeed, there is a fair amount of scatter around the curve — but it can give us a meaningful range of values and a rational basis for a more in-depth analysis and forecasting. Of course, one must not forget that different families of technologies may follow different curves.

Disruptions that Stabilize

But what about “technological revolutions” or “disruptive technologies”? Aren’t they supposed to break all previous assumptions and constraints, and bring totally new, unexpected capabilities? Should we also mention the “revolution in military affairs”? Nah, let’s not go into that perennial ideological swordfight.

A typical evolution of a given class of technology is often described as the S-curve. The progress of a technology begins slowly, then rapidly accelerates, and then slows again to a plateau. Then a different technology — a disruptive technology — emerges, overtakes the previous technology, and goes through its own S-curve. The sequence of such disruptions — multiple S-curves — merge into a roughly continuous curve. One common example is a curve of how the number of computations per second per $1000 of a computing device’s costs advanced from 1900 to early 2000s. This fairly smooth curve proceeds from mechanical calculators through vacuum tubes (certainly a disruptive technology) through transistors (another disruptive technology).

Of course, the S-curves are an oversimplification of reality. But if you squint a little, you may discern similar patterns in the figure at the beginning of this article. You can see how in the mid-1300s muzzle-loading smoothbore firearms started to overcome longbows and crossbows, and eventually plateaued between mid-1600s and early 1800s. You also see how rifle technology began slowly in mid-1500s, rapidly accelerated in mid-1800s, and may or may not yet be facing a plateau about now.

And what about that inflection point around 1830, you ask? It has been noted before for multiple technologies, and probably results from unique and massive changes in the socio-technical history of the mankind: the Industrial Revolution, the American and French Revolutions, and other developments. That’s a very interesting topic for another time.

The point is, disruptions are what keeps the trajectory of technology stable. Without each subsequent disruption, the curve would flatten out. Paradoxically, we need a continuous sequence of disruptions in order to stay on an approximately steady trajectory.

Disciplined Forecasting

To be sure, technological forecasting will never be exact and infallible. Nevertheless, it is an exceptionally important tool for decision-making about major development efforts. It can help reduce the rate of failed programs. It can motivate our persistent, ambitious innovation and keep us from suddenly finding ourselves outranged and outgunned. We must adopt the discipline of continuous, systematic, rigorous technological forecasting. It should be based on historical data, on well-documented methodologies, and on continuous feedback and learning from mistakes. It must not be a one-off initiative, no matter how well-intentioned, but rather a sustained, institutionalized effort.

Such discipline will make sure we pursue audacious innovations without succumbing to fallacies such as the lure of the latest technological over-excitement. We should not rush into way-ahead-of-its-time programs like the visionary but ill-fated Future Combat System. We should have predicted — back in 1990s — that the necessary technology for such a program would not be available for decades yet.

The future is not a silent mystery. It speaks to us from the past, but whispers very softly. We just need to listen more carefully.

Become a Member

Alexander Kott, PhD, is the Chief Scientist of the Combat Capabilities Development Command Army Research Laboratory, a component of the U.S. Army Futures Command. Earlier he served as a Program Manager at DARPA. He has authored over 100 technical papers, and edited and co-authored 10 books. The views expressed in this article are those of the author and do not reflect the official policy or position of the Department of the Army.

Image: U.S. Army (Photo by Christoph Koppers)

Commentary