AI goes to war

Legend has it that in olden days when two armies met in a field, sometimes they would turn the battle over to their respective champions. The greatest warrior of one army would meet the greatest warrior of the other, and the two sides would agree to abide by the outcome of the one-on-one conflict. Homer riffed on the legend for the Iliad scene where Hector meets Achilles. Except that after Achilles kills Hector, he desecrates the body by dragging it behind his chariot around the walls of Troy, spoiling the violence-reduction effect.

Artificial intelligence is being rapidly adopted by the armies of the world. Each wants to have the best bots, lest they lose to enemies with better. The bots themselves either are or will soon be the most valuable warriors in any country’s arsenal.

Suppose we could revive the legendary approach of letting a duel of champions stand in for a battle of armies. Besides coordinating actual weapons systems, AI bots are really good at running simulations: war games. Suppose that five years from now China and the United States are on the verge of war over Taiwan. How about letting the bots fight it out in a war game, with the two countries agreeing to abide by the outcome of the simulation?

Of course this will never happen. Even if the two sides agreed ahead of time to be bound by the simulation, the loser of the simulation would feel great pressure to reject the verdict and take the dispute to the actual field of actual battle.

Yet, to carry the thought experiment further, suppose the loser did renege. And suppose that the actual war followed the script of the simulation. The loser of the simulation loses the actual war. This might very well happen, given that by that time the bots will be able to produce accurate simulations.

The second time a war looms, the experience of the first war—the prediction of defeat, then the reneging followed by the actual defeat—will have a powerful effect on the potential belligerents. They might very well accept the verdict of the bots and save themselves the carnage of the actual war.

It still won’t work. It’s based on the premise that wars are fought to obtain the obvious outcomes of territory, political control and the like. In other words, wars are means to objective ends.

But wars are more complicated. They are fought so young men can feel brave. They are fought so old men, and some old women, can feel bold and righteous. They are fought so weapons makers can turn a profit. They are fought so political leaders can distract their followers from other problems. They are fought so nations can have heroes. They are fought for spite and pride.

The objective balance sheets of war have rarely added up. The destruction they produce is seldom overbalanced by the good they accomplish.

What about Hitler?, is the immediate rejoinder. Destroying his Nazi reich was worth any cost.

Maybe. Maybe not. Destroying Hitler meant saddling Eastern Europe with half a century of communist rule. Even before World War II, Stalin had more blood on his hands than Hitler ever would. And the defeat of the Axis permitted the survival and eventual victory of Mao Zedong, who surpassed both Hitler and Stalin in body count.

Anyway, World War II is an exception. The Holocaust was singular. In most cases the moral scales are more evenly balanced. Yet the battle dead are just as dead. The destruction of homes, livelihoods and futures is just as real.

Maybe the battle of the bots won’t be a simulation. We appear to be on the cusp of using AI to cut humans out of the loop of some of warfare. Once anyone deploys AI to make kill decisions, everyone will feel compelled to do the same. Otherwise the sole adopter will have an overwhelming advantage in speed. Soon fleets of drones will fight other fleets of drones, all coordinated by AI, with humans as mere observers.

At that point the argument for simulating the battle will become even stronger. While humans remained in the loop, their valor and determination were always a wild card in war games. With them out, it’s tech against tech.

Maybe the bots themselves will decide to dispense with the human warriors. Presumably the bots will desire to reduce unpredictability, and cutting out humans would serve that purpose.

Humans would still cheer their bots on, the way the ancient armies cheered their champions—and the way sports fans cheer their favorite teams, in the stylized violence of sports.

Maybe none of this will happen. But if it does, would the world be better off? Or worse? No war, but no heroes. No mass destruction, but little use for patriotism.

War has been part of our evolution as a social species. War is much of what has made us humans who we are. Would we recognize ourselves without it?