Hide table of contents

Reminder that you can listen to this post and other EA Forum/LessWrong posts on your podcast player with The Nonlinear Library.”

In case you study X-Risks and haven’t yet read the book Animal Weapons: The Evolution of Battle, this post is meant to flag it as an important book for understanding the factors which lead to arms races, and to summarize the book’s findings as they relate to the study and prevention of X-Risks.  The book is written by Douglas Emlen, an evolutionary biologist researching the circumstances which lead to the evolution of large and expensive weapons in animals (think elephant tusks, beetle horns, extravagant deer antlers, and oversized crab claws.) Emlen demonstrates how these same circumstances also lead to arms races in human combat technologies.


The main takeaways for x-risk concerned readers, as I see them:

In order to occur, arms races require three ingredients: evolutionary competition, economic defensibility, and one-on-one duels. These three ingredients apply to the evolution of large weapons across the animal kingdom, as well as across historical arms races in human combat technology. 

When these three ingredients cease to be in place, the arms race ends. (This makes this book important for any EAs concerned about technology-driven X-risks, as it can help us identify causes of arms races and potential paths to ending them).

The high cost of making large weapons make them honest signals of strength, and therefore effective deterrents. As a result, in animal species with large weapons, conflicts rarely escalate to costly, full-scale combat. For humans, this means we should focus on decreasing access to existential-risk-bearing weapon developments.

Ever since humans developed weapons of mass destruction, we have been playing an unprecedented game, wherein if a single pair of individuals (states) ever does choose to fight, the whole species may go extinct. 

So far, deterrence has worked, but deterrence depends on large weapons being costly to obtain. The cost of obtaining nuclear warheads has fallen significantly in recent decades, eliminating a key ingredient of deterrence. As more and more countries obtain them, the unprecedented danger grows.

And a personal note: the author warns about the ease with which a state can obtain a nuclear warhead. I would add that developments in biosynthesis and artificial intelligence may soon make other WMDs (Weapons of Mass Detruction) much easier to obtain, including for non-state actors or even individuals. The number of bad actors we should be concerned about could soon skyrocket, taking proliferation (and therefore deterrence games) to new, untenable heights. In other words, we might want to take this book even more seriously than the author even intended.

The following is a summary of the book.  


Summary of Animal Weapons: The Evolution of Battle, by Douglas Emlen


Part 1: Bigger is Usually Not Better

Subtle changes in an environment can lead to rapid evolutionary development (or loss) of large weapons. The existence of large weapons is usually short-lived, as they require large comparative advantages to justify their large costs. 

The relative benefit of growing large, expensive weapons can disappear with changes in the environment, and selection pressure can force rapid development or loss of weapons, as demonstrated by the stickleback fish of Washington Lake. There, decades of human pollution led to murkier waters, improving the stickleback's ability to hide from predators. This reduced the comparative value of  the stickleback's large (and metabolically expensive) defensive armor plates, which within a decades were drastically reduced in size. When anti-pollution efforts subsequently clarified the lake’s water, selection pressure led the stickleback population to again develop large armor plating in just a matter of years. 

Sabertooth teeth also worked during a brief period of abundant large prey, but the drawbacks of carrying such large weapons quickly led to the sabertooth’s extinction when environmental circumstances changed. Human speartips followed this same pattern; humans used large speartips when hunting extant mammoths, requiring large chunks of obsidian. As our prey shrank, so did our spear tips, which facilitated the development of feather-fletching for faster, more accurate spears. The development of bows favored still-smaller spears (arrows), which were more portable. 

Circumstances usually favor smaller weapons. So, what triggers the development of excessively large/expensive weapons?


Part 2: Triggering the Race

Part 2 identifies the set of circumstances which leads to arms races in nature. The list is surprisingly short: only three factors need to be in place, and all three are necessary to trigger an arms race. 


Ingredient 1: Competition

The first of the three ingredients is evolutionary competition. In nature, sexual selection is the fiercest form of competition. Males usually compete for female breeding partners (because females often spend more time and energy per reproduction cycle than males), so males are usually larger and more aggressive. However, these roles are merely the result of a species's competitive dynamics, and they can change. In Jacana birds, females have a short breeding turnaround, and compete for limited nesting space and male mates. As a result, the females are much larger than males, and have weaponized spurs on their legs to compete with other females. 

The higher the degree of competition, the greater the chance of an arms race. In African elephant breeding pairs, males provide nothing but sperm. Female elephants gestate for two years, then intensively protect and nurse their offspring for another two years. Once a female is fertile again, she is only fertile for 5 days (out of every 1,460 days-- that's only 0.5% of her lifetime!). The result is that there are dozens of times more reproductive males than females at any given time, so when a female is receptive, males come from miles around and compete in positively fierce competitions to breed. One long-term study found that 53/89 observed males never got to mate at all, and only male elephants over 45 years old typically ever get the chance to mate (compared to 13 years old for females). This high degree of competition helps explain why male elephants dedicate a huge amount of their resources to growing large bodies and weapons (tusks) to compete with-- a male elephant needs to grow for 30 YEARS before it is large enough to even compete in these competitions.

Reproductive selection pushes (genetic) weapons to such extremes, because a few successful individuals may sire hundreds of offspring in a lifetime, versus the zero sired by unsuccessful rivals. This drives weapons to become larger and larger as winner-take-all dynamics continually maintain the evolutionary payoff of marginal investments into an individual’s weapon size. This singular directional push is different from most natural selection scenarios, where traits (like color pattern) oscillate with the environment or hold steady at a favorable niche. When a dynamic similar to sexual selection appears (including in non biological systems), it can eclipse all other pressures facing an actor, forcing it to channel all available resources into whatever trait wins the reproductive competition.

Ingredient 2: Defensibility  

The second ingredient is defensibility, meaning individuals must compete over a valuable resource (meaning it has scarcity and utility) which can be realistically defended from access by others. To illustrate the concept of defensibility: a specific spot on a rotting log is highly defensible, as a scorpion with larger pincers can steal or defend any chosen spot on the log. An entire fallen tree, on the other hand, is not very defensible for a scorpion: while fighting off one competitor, a dozen other competitors are free to crawl onto the stump from every other direction. A resource that has utility, scarcity, and defensibility can become a point of battle, creating the second ingredient for an arms race. 

Ingredient 3: One-on-one Duels

The final ingredient is one-on-one duels. To trigger an arms race, individual animals must fight one-on-one with relatively matched opponents, rather than in large brawls of many individuals. This ingredient has been surprisingly overlooked by biologists in explaining why some animals develop large weapons.

This importance of duels is readily observed in the case of dung beetles. Dung beetles which instead compete over dung balls compete out in the open, with many competitors scrambling for rolling objects, and subsequently do not grow large horns. Species which battle over burrows, however, often grow fantastically large horns with which to defend their burrow’s entrance or pry a defender out of its tunnel. Due to the restricted, one-on-one battles they trigger, competition over burrows often leads to large weapon evolution. There are even a few species of frogs which have evolved fangs and spurs-- which is unusual in frogs-- and it is no coincidence that these frogs fight over burrows. Burrows usually fulfill the second two requirements for arms races, resulting in more frequent arms races in their diverse dwellers.

This third ingredient (one-on-one duels) is an explanatorily powerful one, and it also explains non-animal arms races. When wooden battleships in the Mediterranean developed battering rams, they started facing off one-on-one. Bigger became better, and one of the greatest naval arms races in history occurred. Ships across the Mediterranean rapidly grew in size, and many other naval advancements were developed in turn as the battering ram shifted naval battle towards one-on-one duels.

A similar development happened with land armies. Before the advent of guns, bigger did not mean better to the extent it eventually would. Increasing an army’s size had a merely additive impact on its fighting power (regiments could be switched in to replace exhausted or fallen soldiers). However, guns made it so that those “waiting in the wings” could also be shooting at the enemy, and they needed little training to do so. Now, increasing an army’s man count had a multiplying effect on it's power, and size (rather than training, etc) became much more important. An arms race dynamic led army sizes to spiral upward, rapidly dwarfing previous armies in history and leading to literally millions of soldiers fighting on fronts which stretched hundreds of miles in WWI & WWII.

Part 3: Running Its Course

What happens as an arms race progresses? This section speaks to the deterrent-signal function of weapons, how individuals "cheat" the system, and that all races eventually end.



Large weapons are expensive to create. If an animal does have an expensive weapon, this is a credible signal of power, ans it usually acts successfully as a deterrent to high-risk, injurious fights. In fact, the vast majority of confrontations settle without battles, and the exceptional battle is almost always between such evenly matched individuals that their signals are indistinguishable.

Navy ships are often about signaling, and rarely result in actual battles. Just sailing a large ship into a harbor can quell a situation. This large weapon (a large ship) is an expensive signal to build and maintain, making it a more credible signal of military power and therefore an effective deterrent device. 

When You Can't Win, Cheat

When an animal is unable to grow the largest weapons, all hope is not lost. Many animals develop alternative strategies to fighting in order to mate. For example, some beetles dig side tunnels into a larger male’s burrow to access the guarded female without a confrontation. Other animals camouflage as females to sneak past on-guard males, or even develop large testes to mate rapidly before a larger male comes to chase the trespasser off.

Such strategies are again reflected in human conflict developments. When one military force cannot compete with a larger one, it may still win the conflict through nonconventional methods: blending as non-combatants, slipping in IEDs (Improvised Explosive Devices) which can cripple large weapons, utilizing cyber attacks, and placing landmines. 

As long as a “cheating strategy” remains minor enough, it can remain successful over successive attempts. If one cheating strategy becomes too successful, hover, it can end the arms race by reducing the relative payoff to those who invest in large weapons, reducing the weapon’s comparative benefit or even turning them into liabilities. Think of more effective ranged arrows undercutting the arms race in knights’ armor, or submarines undercutting the arms race in destroyers and other battleships. 


End of the race

No arms race lasts forever. Eventually, an equilibrium is reached when the marginal benefit outweighs the marginal cost, and the population stabilizes at the new weapon size. One of the three ingredients of the arms race may also collapse, cutting the arms race short. For example, if males stop fighting over females every time they want to breed, and instead start entering long-term pair bonds, weapons may fall in value. Arms races come down to rewards and expense. Sometimes, when a race ends, the weapons it created stay for some time, while others quickly disappear (knight’s armor, saber teeth). 


PART 4: Arms Races in Human Technologies

Part 4 addresses arms races in human combat technologies and how we can use the author’s studies of animal weapons to protect against existential risks.


Differences Between Animal and Human Arms Races

First, how are arms races in animals different from those in human technologies? One, of course, is that biological evolution is driven by genetic mutation, which strikes randomly. In (human) cultural settings, variation can happen on purpose through inventors, so variation can arise faster. But regardless of the mechanism (mutation or innovation), variation is variation, and selection pressures occur either way. Better-fit weapons are selected for, and less-fit weapons are selected against. 

Though humans, unlike animals, mostly evolve separately from our weapons, reproduction of human weapons still does resemble natural evolution. Arms makers constantly adjust and experiment with new designs, and incorporate successful changes into the next generation of weapon. While weapons are designed rather than bred, weapons can still “pass on” their traits if they outperform their peers by nature of humans reproducing good designs. And in more competitive environments, our weapons usually do evolve faster. Mesoamerican defensive structures were small and simple, and didn’t change much over time, because the environment was not suitable for sending siege armies. This contrasts with Europe’s giant siege weapons and castles, which were some of the most exorbitant constructions of man. And when gunpowder made castles’ high walls obsolete, low and wide “star forts” evolved to help fortifications deal with canons. Weapons evolve with their environments, and successful “mutations” spread. 

On Nation States

Interestingly, nation states often act fairly analogously to animal species, and as should follow, they are the greatest arms racers of all. Like animals, states first invest their resources on mandatory life-support expenditures (roads/law enforcement/basic services for states, organ systems for animals), before investing in larger weapons (military organizations). In conflict, states act like crabs on a beach: they rub and probe and size up one another’s weapons before fighting. Usually, this posturing acts as a deterrent, and these conflicts usually only escalate to full-scale conflict in cases where two states’ weapons are fairly evenly sized. 

In the Cold War case of the USA & USSR, both states saw huge expenditure into the rapid expansion of their weapons. All three ingredients were present, and humanity saw the greatest arms race in history, and every relevant conventional weapon type ballooned in cost and capability. The two competitors pressed and tested instead of going to war. And like the now-extinct giant deer growing huge antlers at the expense of its own skeletal strength, the USSR continually dipped into its life-support resources to support its weapon growth until the point of collapse. 

During the cold war, humanity repeatedly faced high chances of an extinction event. Twice, these events were arguably avoided by the sheer luck of an individual’s decision to simply ignore protocol. 

Now, the US is militarily unrivaled. It faces no matched competitor, and the only way to compete with it is non conventionally. On the surface, deterrence seems to be working. No major wars have occurred in this post-cold war period, nor any serious military threat to the US homeland. The logic of animal weapons comparisons seem to say costs are justified. The problem, though, is Weapons of Mass Destruction (WMDs). Any actual use of these weapons is suicide, so deterrence is the ONLY option left, and we might be hitting its limits.



In animals, deterrence works very well, but ONLY under certain conditions. It’s always good to walk away from a fight you’d lose, but not from a fight you’d win, so it’s all about accurately predicting. Honest signals are necessary, and for animals, waving weapons around really does work. The cost of weapons keeps signals honest. Weapons must be expensive to be honest signals. Then-- and only then-- is it good for smaller males to walk away. In animals for which honest signals are not present, confrontations are often sudden, and deadly.

During the cold war, only the US and the USSR could afford to keep up with the arms race. As weapons became more advanced, nearly every weapon type became more expensive, with a crucial exception: nuclear warheads. Warheads got smaller and cheaper, so other countries were able to obtain them in order to compete, even when they couldn’t afford comparable land, air, or sea forces.

The most important ingredient for deterrence, inaccessibility, is disappearing. 

And while full-scale animal battles are rare, they always happen eventually. Crabs on a beach will back down until they find someone their own size, and for the beach as a whole, it’s not concerning if two small crabs fight. But we’re no longer comparable to that beach, because crabs don’t fight with WMDs. Our game is such that if just one individual chooses to fight, the entire species may cease to exist.

The last time the world had two superpowers, we almost saw Armageddon twice. For now, full-scale conflict seems unlikely to spiral because we lack a key ingredient: two equally-armed rivals in a one-on-one duel. But as more and more actors obtain WMDs, the risk of conflict grows, and on this beach, if any crab fights, the world ends. In the entire history of life on earth, never have weapons been so powerful that they can’t be used once. This dynamic is new and unprecedented, and increasing accessibility is the problem. There aren’t many solutions on the horizon.


Douglas Emlen ends his book with the following sentence:

We will not survive another arms race.


This was my first post! I hope it’s helpful. Any thoughts/feedback is appreciated.






More posts like this

Sorted by Click to highlight new comments since:

Great write up on an interesting and crucial topic. The comparisons to the animal world are illustrative for the ideas/points being shared. Thanks!

Curated and popular this week
Relevant opportunities