A Theory of Equilibrium in the Offense-Defense Balance
Don't Make the Peak Oil Mistake When Thinking About the Offense-Defense Balance
The offense-defense balance is a concept that compares how easy it is to protect vs conquer or destroy resources. For example, autonomous weapons and AI systems might make attacks easier and more scalable compared to defensive measures. The balance matters because when offense has the advantage, it can create instability and increase the likelihood of conflict.
I claim that worries over a massively upset offense-defense balance make the same mistake as peak oil apocalypse prophecies.
Peak Oil
Peak oil hysteria was seeded in the environmentalism and energy crisis of the 1970s and reached its apogee in the early 2000s with a flood of books, documentaries, and movements predicting civilizational collapse.
In some ways, the predictions of peak oil prophets did come true. Oil prices rose to an all-time peak in 2008, and oil consumption in the US reached its peak around the same time. Few, if any, of the forecasted societal or economic collapses came to fruition though.
The problem with the peak oil predictions was a selective extrapolation of some trends, while holding the adaptation of others constant. They projected rising oil prices and squabbles over the remaining finite supply that would force more and more uses for oil to shut down, but they didn’t project the new substitutes and oil sources that these high prices would create.
When oil prices rise from $100 to $200 a barrel, it doesn’t double everyone’s energy costs. $200 oil prices don’t force all plastic manufacturers to pay $200 for their chemical inputs, for example. They force them to substitute petroleum and use $150 dollars of vegetable oil instead. The new $200 price is an upper bound on what previous oil consumers pay after the price increase. Any consumers with substitutes that cost less than $200 will use those instead. So even in the most extreme world where oil runs out, most oil consumers aren’t faced with infinite costs and forced to shut down. Instead, they are only forced to pay for their next best substitute.
Offense-Defense Balance
As an example of how this mistake applies to the offense-defense balance, consider the offense-defense balance of presidential assassination. In the 19th century, guns were inaccurate at long ranges, so assassins needed to get close. Security used this defensive advantage to protect the president, but some assassins did slip through. Accurate, long range rifles are a massive upset to the offense-defense balance of presidential assassinations. Instead of needing to sneak up on the president in a theatre, you could sit half a kilometer away and take shots from there. These accurate guns are cheap and accessible.
But it would be a mistake to predict some large increase in the rate of presidential assassination. That prediction only makes sense if you extrapolate one trend in offensive capabilities, but hold everyone else’s adaptation constant. A rise in the offensive capability of assassins doesn’t force the secret service to accept a higher rate of presidential assassination. They are just forced to buy The Beast and put them behind glass.
Applying this to the future, consider drone-powered assassination. Assassin drones with bombs strapped to them will be small, fast, cheap, and potentially autonomous. This will make it easier to attempt to assassinate world leaders, but again it would be a mistake to project a new era of instability and terrorism based on this change. The death of a world leader is extremely costly. So any adaptation that the secret service can use to neutralize drone assassins that’s cheaper than letting more presidents die will be used. The effect of this shift of the Offense-Defense balance won’t be more deaths, it will be investment into powerful EMPs or counter-drones, or good old bullet proof glass.
The usual extrapolations of the offense-defense balance you hear are upper bounds of the costs that new offensive technologies can impose on defenders. If new technology enables $10,000 of drones to destroy a $10,000,000,000 dollar aircraft carrier, the actual cost imposed on defenders will not be losing all of their carriers. As long as there is any investment they can make that costs less than $10,000,000,000 dollars and neutralizes the drones, they will make it. Thus, shifts in the offense-defense balance are attenuated and somewhat self balancing: the more value that is imperiled by a new offensive technology, the more options that defenders can afford to neutralize it.
When the offense-defense balance changes due to some technology, prices and investment adjust to maintain a more stable equilibrium.
Specifically in the military realm, there's also a speed issue. During actual war, everyone pulls out all the stops and stuff develops at breakneck speed. But during peacetime, there's no need to do that. So we war-game out what we'll need, and leisurely work on that, and then when war happens, we go to war on day 1 with the military technology we have. But, you can, in all-out war, move fast enough to pre-empt your opponent's wartime adaptation.
The question is, what if one side develops a sufficiently cheap/effective offense (say, based on the war-time development incentives in a small, regional conflict) that can be deployed quickly enough not to just win the first battle, but destroy the capability (either by destruction of resources/factories, or political will) of the other side to restore the balance?
"As long as there is any investment they can make that costs less than $10,000,000,000 dollars and neutralizes the drones, they will make it."
But investments also take time and political capital. If the 10,000 drones wipe out a carrier group, the domestic political damage will exceed even the dollar price tag. "Too big to fail" can apply to more than just banks.
I'd add that, in practice, states will mostly value weapons systems such as aircraft carriers instrumentally--for acquiring or protecting intrinsically valuable interests--and if so, the relevant threshold for a carrier-protecting investment is the value of the intrinsic interests at stake, rather than the cost of the carrier. If a state spent $10bn on a carrier in order to protect a far-flung island chain that it values at $15bn, and a breakthrough drone technology renders the carrier vulnerable to cheap destruction, the state should invest in an effective anti-drone system for the carrier if it costs less than $15bn (or, if possible, find a cheaper way of protecting the islands).