America’s Next Soldiers Will Be Machines

FORT IRWIN, California—For as long as the United States has had an army, U.S. infantry soldiers have stuck by one motto: “Don’t shoot until you see the whites of their eyes.”

FORT IRWIN, California—For as long as the United States has had an army, U.S. infantry soldiers have stuck by one motto: “Don’t shoot until you see the whites of their eyes.”

But fighting on the U.S. Army’s largest training ground last month, Lt. Isaac McCurdy and his platoon of infantry troops, playing a fictional enemy of the United States, found themselves up against a very different kind of foe: one with camera lenses for eyes and sheet metal for skin.

These weren’t your average flesh-and-blood men that they were fighting. They were machines.

Driving on eight screeching wheels and carrying enough firepower on their truck beds to fill a small arms depot, a handful of U.S. Army robots stormed through the battlefield of the fictional city of Ujen.

The robots shot up houses where the opposition force hid. Drones that had been loitering over the battlefield for hours hovered above McCurdy and his team and dropped “bombs”—foam footballs, in this case—right on top of them, a perfectly placed artillery shot. Robot dogs, with sensors for heads, searched houses to make sure they were clear.

“If you see the whites of someone’s eyes or their sunglasses, [and] you shoot back at that, they’re going to have a human response,” McCurdy said. “If it’s a robot pulling up, shooting something that’s bigger than you can carry yourself, and it’s not going to just die when you shoot a center mass, it’s a very different feeling.”

A robotic dog moves across a sandy desert landscape through the purple smoke of a flare. Solders holding guns crouch behind the four-legged robot, taking shelter against a wall.

A Ghost Robotic Dog moves forward with U.S. soldiers behind it during an exercise at Fort Irwin on March 17Spc. Samarion Hicks/U.S. Army

The Army’s top officer, Gen. Randy George, made the trip from the Pentagon to watch the fictitious battle. Standing on the rooftop of a makeshift stucco building that swayed from the tank rounds thudding against the desert sand, he was flanked by a half-dozen top generals and about as many colonels. Many of them wore metal wristbands engraved with the names of the soldiers who died under their commands in Iraq and Afghanistan.

In the United States’ next major war, the Army’s brass is hoping that robots will be the ones taking the first punch, doing the dirty, dull, and dangerous jobs that killed hundreds—likely thousands—of the more than 7,000 U.S. service members who died during two decades of wars in the Middle East. The goal is to put a robot in the most dangerous spot on the battlefield instead of a 19-year-old private fresh out of basic training.

Not all of these machines are ready for prime time, though. During the Fort Irwin exercise, dubbed “Project Convergence”—which was designed to test robots in realistic combat scenarios—they didn’t face live bullets. The robots don’t have peripheral vision; they can’t look left or look right like a human soldier can by simply turning their head. And the Army’s outdated network can’t always keep hundreds of drones aloft at the same time, or even tell U.S. troops which of the unpiloted aircraft are friend or foe.

But they’ve come a long way. George and other Army leaders believe that almost every U.S. Army unit, down to the smallest foot patrols, will soon have drones in the sky to sense, protect, and attack. And it won’t be long before the United States is deploying ground robots into battle in human-machine teams.

“It’s not just one robot [that] replaces one human,” said Alex Miller, George’s chief technology officer. The question, he said, is how the Army can put robots on the front line “without taking a single rifleman, or multiple riflemen, off the line to control a robot.”

A pile of boulders in a desert landscape are painted all over with U.S. Army logos in various colors. A blue sky and a few small white clouds are visible overhead.

U.S. Army Futures Command and Project Convergence logos are painted alongside various insignias on rocks outside the gates of the National Training Center at Fort Irwin, seen on March 20.Spc. Marquis McCants/U.S. Army

You could be forgiven if you didn’t know there was a U.S. Army training base nearly the size of Rhode Island in the middle of the California desert. After all, Fort Irwin isn’t easy to get to.

Pull off Interstate 15, go through the city of Barstow, and drive another 23 or so miles. You’ll see a big pile of rocks—boulders of basalt formed millions of years ago when an ancient volcano erupted—painted with Army insignias. You’ll see the red and blue shield of the 197th Infantry Brigade and the black and yellow Norman shield and horse silhouette of the 1st Cavalry Division. Every logo symbolizes a different Army unit that has trained here—about 10 every year. That’s how you know you’re at Fort Irwin.

The isolation is the point. Soldiers call the massive training range “the box.” That’s because it’s basically a sandbox for military exercises, one of the only places in the continental United States where you can jam the electromagnetic spectrum, fly drones at 30,000 feet, and fire large munitions without bothering anybody. Fort Irwin contains 14 fictional cities and 1,200 buildings. Once troops arrive, they’re here for 21 days—with no showers.

Life in the desert is hard. In a January exercise, U.S. Army troops had to brave wind, rain, sleet, snow, and oppressive heat, sometimes all in one day. There are about 1,200 fake casualties for every three-week rotation. Service members in one of the brigades that deployed here for that mock battle with the 1st Armored Division had to carry their artillery with them 40 miles into battle.

Soldiers in combat gear run across a dusty street with their guns drawn in front of the simple concrete buildings of a fake city used for training. Telephone wires and clouds are visible overhead.

Soldiers conduct an urban assault drill in the training grounds at Fort Irwin on March 18. Sgt. Maxwell Bass/U.S. Army

It’s like a giant game of laser tag, with the intensity turned up. Almost everyone who fights in the simulation dies in the simulation at one point—sometimes with pools of fake blood and prosthetic gore that looks real enough to fool an untrained eye. Even George has died in the simulation more times than he’d like to admit.

The scenarios are written to fit wherever U.S. troops are deploying, which for the past two decades has mostly been Iraq and Afghanistan. But since 2014, around the time of the Kremlin’s first invasion of Ukraine, the Army has used Fort Irwin to practice large-scale combat operations against enemy forces modeled after China’s People’s Liberation Army and the Russian military.

The idea is for the Army to do most of its bleeding in training, not on the battlefield. So organizers try to make the simulated enemy tougher than the real one—not like the Russians, who stumbled into Ukraine with maps from the 1980s looking for towns that didn’t exist. The fighting that Ukrainian troops have experienced on the battlefield—inside Mariupol’s Azovstal steel plant, around the city of Kharkiv, and during the liberation of Kherson—has forced the Army to adapt even further.

“Originally, we had units actually knocking on doors because that’s what they learned in Iraq and Afghanistan,” said Maj. Gen. Curt Taylor, the commander of the national training center. “That’s going to get you killed.” And more often than not, soldiers have to hold their (simulated) fire. There are paid actors roaming around Irwin playing civilians caught in the crossfire.

An 8-wheeled unmanned vehicle with a gun on the front moves across a sandy street as soldiers in combat gear watch.

An unmanned transport vehicle with a remote weapon system provides support to soldiers in an urban environment as part of a training exercise at Fort Irwin on March 12. Spc. Samarion Hicks/U.S. Army

The scenarios are also starting to include robots. Today, the Army might have three platoons deployed in a basic attack, McCurdy said: one to fix the enemy in place, another to maneuver on them, and then one more in reserve. Put robots into that group of soldiers, and they suddenly have much more room to maneuver.

“Then you have basically three platoons entirely freed up,” McCurdy said. “And it also draws far more enemy attention to those harder targets of the robots. It protects you, frees you up, gives you more flexibility, and weakens up all other fronts of the enemy’s defenses.”

“I’m not 100 percent sure how they would work against it,” said Sgt. Philip Webb, the weapons squad leader for the opposing force in the exercise.

That might not be true on every battlefield. Like everything here, Project Convergence is an experiment. It’s hard to stop a robot from rolling along Fort Irwin’s sand and volcanic rock, which has plenty of traction for the Army’s Small Multipurpose Equipment Transports to move freely across. And soldiers haven’t shot any real live ammunition at the robots—yet. A well-placed shot to the sensor might render a ground robot functionally useless. And robots might turn into expensive icicles at the British military’s largest training ground in the hostile winters of the Canadian province of Alberta.

Drones are already ubiquitous—you can buy a quadcopter off the shelf for about $50 at most big-box stores—but building ground robots that can see and sense their way through modern battlefields is much harder.

But the entire field of military robotics is speeding up. Ukraine has already pledged to create a separate service branch for drones and unmanned vehicles. The U.S. Army might not be too far off: It’s considering a proposal to add a platoon of robots, the equivalent of 20 to 50 human soldiers, to its armored brigade combat teams, which are the service’s bread-and-butter, tank-backed infantry units.

Army officials don’t think that robots will replace humans—or that they will do much to solve the problem of recruiting shortfalls—but they’re working on the math to get the ratio of humans to robots to about 2-to-1 or 3-to-1.

The point is to get the advantage before China or Russia do.

“We don’t want it to be even,” said Lt. Gen. John Morrison, the principal advisor to the Army’s chief of staff for network and cybersecurity. “We want it to be a technical overmatch.”

A soldier in combat gear looks up as a small drone flies above him in front of a pile of rocks in a desert landscape. Another soldiers kneels in front of a massive pile of boulders a few yards away.

Soldiers use experimental drone technology in a training assault as a part of Project Convergence at Fort Irwin on Oct. 27, 2022. Spc. Jaaron Tolley/U.S. Army

If the Russia-Ukraine war has taught the Army anything about the future of warfare, it’s to look up. Soldiers are listening for drones and coming up with battle drills to defend against them. If something can be seen or sensed, it can be killed.

To survive on the modern battlefield, soldiers are having to make themselves smaller and smaller—almost invisible. Loitering munitions can wait over the battlefield for hours, ready to dive if an operator senses the faintest twitch. Fort Irwin’s mock enemy troops have 105 drone swarms and can attack with dozens to hundreds of the small unmanned planes at a time.

Unwieldy server stacks and satellite dishes sticking out of U.S. vehicles and outposts could get soldiers killed. And U.S. troops will probably have to turn off their iPhones and Androids, which are rippling through the electromagnetic spectrum.

“The cell phone is the new cigarette in the foxhole,” said Taylor, the training center commander. “We’ve shown soldiers, ‘hey, your cell phone can get you killed.’” Taylor and his team recently spotted an otherwise undetectable Apache helicopter weaving through their air defenses when they clocked the pilot’s iPhone doing 120 miles per hour across the desert.

War is moving at the speed of technology—light-years. But the U.S. Army moves at the speed of budget—fiscal years. That’s not going to work when a home-built quadcopter can drop a grenade and destroy a fighter jet parked on the ground. And the Defense Department is on the wrong side of the cost equation.

“What keeps me up is the cost imposition factor that robotics systems bring,” said Miller, the chief technology officer. “A $1,500 drone can take out a multimillion-dollar aircraft.”

A U.S. Army general in camouflage wears an augmented reality headset over his eyes and holds a controller. Standing around him are reporters and other soldiers.

Gen. Randy George tests new augmented reality technology during an event at Fort Irwin on March 18.Sgt. Maxwell Bass/U.S. Army

George has only been on the job for six months, but he has already taken a scalpel to the Army’s weapons programs. The Army canceled the RQ-11 Raven and RQ-7 Shadow drone programs, which took nearly a decade to produce and earned the nickname “flying lawnmower” from troops miffed by their slow speed. Crash just one and it might cost the Army a million dollars—enough to pay about a platoon’s worth of Army privates their full year’s wages.

But building up is difficult, and China is poised at virtually every part of the drone and robotics supply chain. And Project Convergence is still just an experiment; the Army hasn’t been able to get most of this new technology into the budget—it’s mostly on the list of unfunded priorities that George is asking Congress to add to its almost $186 billion budget request. That’s too slow. In the next war, the Army is going to need to reprogram things instantly.

With drones becoming all the rage, the Army has also canceled the big-ticket Future Attack Reconnaissance Aircraft program—which was developing a next-generation scout helicopter—and service leaders want to end production of one variant of Blackhawk helicopters. It’s still an open question whether the Army has the authority to buy and build cheap off-the-shelf drones like the Ukrainians, though. George is seeking the acquisition authority to do so, but the defense industry doesn’t love the idea of the Army being a less predictable weapons buyer; many companies want the stability of multiyear contracts.

The next target for belt-tightening is the ponderous Distributed Common Ground System, the Army’s most widely used intelligence sharing platform: stacks of servers that resemble discarded record players with messes of wires extending from them, many of which have had to be updated by hand.

Members of the brass say that they’re listening to the rank and file’s demands to throw out the old junk. “When someone says something is a piece of crap, that bothers me,” George said.

Instead, troops want to be running their missions from tablets and iPhones. Maj. Gen. Jim Isenhower, who leads the Army’s 1st Armored Division, told George during a demonstration that he believes putting command-and-control systems onto screens, tablets, and mixed reality applications can help him reduce the size of a combat outpost by 95 percent, from 326 troops to about eight—so small that the Army’s opposition force didn’t even spot it during practice drills.

Calling out orders from a tablet can make the Army move faster, too. Before, U.S. Army intelligence analysts were looking at radar and signals intercepts through a soda straw, trying to find the right line of data to see what weapons the enemy had and where they were. Now, with one radar pass, U.S. troops can start to map out a Russian and Chinese unit and what kinds of guns that unit has in its arsenal, with artificial intelligence helping to find the signal in the noise.

The goal is to reduce the speed of targeting decisions to minutes and seconds in order to compete with enemies who are equally fast, if not faster.

“You defend against the first wave, great,” said Col. Scott Shaw, the director of the Army’s maneuver capabilities development and integration directorate. “What about the 12th wave? And what if the 12th wave comes in the 12th minute?”

Two soldiers stand and one lays on the ground, all in combat gear, with guns drawn, as they navigate a street corner on a simulated streetscape during a training drill in the desert.

U.S. Army paratroopers observe their surroundings during an urban terrain drill at Fort Irwin on March 11.Sgt. Maxwell Bass/U.S. Army

Plumes of smoke drifted off the desert sand and the robot’s gun barrels. The SMET vehicles that had stormed through Ujen had wheeled into position and fired Javelin shots into the caldera, the valley’s cauldron-like floor.

In their wake, mine-clearing robots rushed out and fired a 350-foot rope line full of explosive charge, more than 1,500 pounds of C-4 explosives spooled in a tube.

“Five. Four. Three. Two. One,” Shaw said, counting down as he waited for the charge to detonate.

It didn’t go off. So Shaw counted down from five again. Again, there was no explosion.

The faulty line charge was one of the few things that went visibly wrong during the exercise. In another instance, Army officials jammed themselves, and a swarm of drones dropped out of the sky. If soldiers were to make these kinds of mistakes on a real battlefield today, actual human beings could get killed.

“I would not want to put a human into that,” said Taylor, Fort Irwin’s commander, who has watched videos of Ukrainian troops trying to clear Russian minefields with the same U.S.-made explosives, working under a hail of gunfire.

Robots may be harder to kill, but you can’t fix stupid, either. “They don’t have curiosity—they don’t have instincts,” said Gen. Jim Rainey, the head of U.S. Army Futures Command.

An unmanned vehicle assisted by soldiers in combat gear is loaded with human-sized dolls used to represent fake casualties during a drill. Another vehicle is at right with low buildings behind.

An unmanned military transport vehicle deploys for a simulated casualty evacuation during an exercise at Fort Irwin on March 16. Spc. Marquis McCants/U.S. Army

But the idea of creating robots that do have curiosity and instincts is also terrifying. Are military officials worried about creating terminators?

George, the Army chief, said that he’s not. “You’re talking about completely autonomous vehicles,” he said. “Everything out there was controlled by a soldier.”

Experts acknowledge that the robots of the future may need two sets of rules: one for peacetime, and another for war. With Russian troops invading their country, the Ukrainian military doesn’t care very much about safety features such as kill switches, which would force a robot to cease all operations if it gets a command that’s not signed with the encryption codes of the right controller.

The United Nations, on the other hand, cares a lot about such things. The U.N. General Assembly’s First Committee approved a draft resolution on autonomous weapons in November, declaring that a human should always be in the loop.

“The law of warfare is how we’re going to apply robotic systems,” said Miller, George’s chief technology officer.

But scientists are still at the ground level of building out the field of robotics, just like the Stanford University professors who built the network protocols that led to the internet. “They didn’t envision the internet, so they didn’t envision all the ways that people could abuse the internet,” said Miller.

“The scariest circumstance is not Skynet,” Miller added, referring to the super-intelligent network of killer robots from the Terminator movies as he held a cellphone with an app tracking the positions of soldiers and robots at Fort Irwin. “The scariest circumstance is that it comes under the enemy’s control.”

Read More

Jack Detsch