When the Pentagon Subscribes to the Future

It has long been a staple of the more imaginative sort of fiction that the wars of the future would be fought by gleaming, slightly judgmental machines with a penchant for dramatic lighting. However, as the US Army has recently demonstrated by handing twenty billion dollars to a company named Anduril, the reality is far more likely to involve a very large software subscription and a series of increasingly complex user manuals.

The contract in question concerns a system known as Lattice. One might be forgiven for assuming that Lattice is a particularly robust variety of garden fencing or perhaps a new method for weaving artisanal sourdough. In the hands of the Pentagon, however, it is an AI-powered platform designed to connect everything from drones to ground sensors into a single, cohesive digital consciousness. It is, in essence, the world's most expensive and high-stakes group chat.

There is something quintessentially modern about the idea of 'software-defined warfare.' It suggests that the primary challenge of the modern battlefield is not so much the physical presence of the enemy, but rather the difficulty of getting one's various gadgets to talk to one another without the digital equivalent of a 'spinning wheel of death.' One imagines a general in a bunker, squinting at a screen and wondering if the reason the autonomous drone isn't firing is because it's currently installing a critical security update for its navigation sub-routine.

Anduril, named after a legendary sword that was broken and then reforged, seems to have taken the 'reforging' part quite literally, though they have replaced the blacksmith's hammer with a series of very fast processors and some rather elegant code. The company's approach is to treat the battlefield as a data problem. If you can just get enough sensors to agree on what they are looking at, the theory goes, the fog of war will lift, revealing a perfectly organized spreadsheet of targets and tactical opportunities.

Of course, the transition from human-led intuition to algorithmic certainty is not without its whimsical little hiccups. There is the persistent question of what happens when the algorithm develops a particularly strong opinion about something that a human might find trivial. One can almost hear the polite, synthesized voice of the Lattice system explaining that it cannot possibly authorize a strike on the enemy's fuel depot because the lighting is suboptimal for its computer vision model, and it would much prefer to wait for the golden hour.

Fictionalised Observation: I once saw a man attempt to use a voice-activated toaster to prepare a crumpet. The toaster, convinced he was asking for a weather report for the Greater London area, refused to engage its heating elements until he had confirmed his postcode. I suspect the Pentagon's new AI might have similar moments of bureaucratic stubbornness, albeit with significantly higher stakes than a cold crumpet.

The scale of the investment—twenty billion dollars—is the sort of number that usually requires a few moments of quiet contemplation and perhaps a stiff drink. It is enough money to buy a small country, or at the very least, a very large number of very high-quality pens. That it is being spent on a digital 'lattice' suggests that the future of defense is no longer about who has the biggest boots on the ground, but who has the most efficient cloud storage.

There is also the matter of the 'autonomous' part of the equation. While the military is always careful to mention that there is a 'human in the loop,' one can't help but feel that the loop is getting increasingly large and the human is getting increasingly tired. It is a bit like being the designated driver for a car that is perfectly capable of driving itself, but insists on you holding the steering wheel just so it doesn't feel lonely.

In the end, the Anduril contract represents a fundamental shift in how we perceive the machinery of conflict. We are moving away from the era of the heroic pilot and into the era of the heroic systems administrator. The medals of the future may well be awarded for 'Outstanding Achievement in Latency Reduction' or 'Gallantry in the Face of a Corrupted Database.'

Fictionalised Observation: My uncle, a man who once tried to fix a television by shouting at it in Latin, would have found the idea of a 'software-defined' army entirely incomprehensible. He believed that if you couldn't hit it with a spanner, it wasn't real. I often wonder what he would make of a twenty-billion-dollar fence made of light and logic.

As we watch the Lattice spread its digital vines across the landscape of modern defense, we must hope that the gardeners in charge have a firm grasp on the pruning shears. After all, the only thing more dangerous than a battlefield you can't see is one that is being managed by an algorithm that has decided, quite reasonably, that the most efficient way to win a war is to simply refuse to participate until the budget is increased.