P.W. Singer is a well-known figure in the defense community. A staple in the think tank crowd, he wrote Wired for War, the seminal public text on robots in combat, and Ghost Fleet, a speculative World War 3 novel that led to him testifying in front of the Senate Armed Services Committee. He's served as a technical consultant on games in the Call of Duty and Metal Gear series, as well as Hollywood films. When serious people want to talk about machines in warfare, they call Singer, or visit him at his office at the New America Foundation, a block from the White House.
So when a military wargaming project called him in to consult on a defense simulation trying to predict the wars of tomorrow, it was par for the course. But when he got there, he found that they'd used images from sci-fi films like The Terminator to represent what battlefield robots would look like.
"Why aren't you pulling images of real-world ones that are operating right now?" he asked. As surprising as it might seem, even military planners default toward sci-fi robots when they imagine the future of war. "My role has been to say: Don't pull completely from mechs."
Yet despite the wealth of real-life military robots, games about future war tend to default to sci-fi as well. Just in the last month, two games about future conflicts—Call of Duty: Infinite Warfare and Titanfall 2—imagine a world populated by autonomous war machines that mimic human anatomy, from anthropomorphic robot infantry to AI-governed mechs.
How realistic is this, exactly? Does this square with how robots are actually used on the battlefield, and would you even want to have a robot shaped like a person, with arms, legs, and a head? It's a surprisingly complex question, and reveals an incestuous middle ground between the future robots artists imagine and the very real hardware scientists are building.
Despite sci-fi overtones, the question of what military robots will look and act like is shockingly relevant to the present day—because next year, the Trump administration will decide whether robots should target and kill without human input.
Taking It Step-By-Step
The contrast between practical robotics and science fiction has existed since the birth of the genre. In "Robbie," the opening story of Isaac Asimov's 1950 collection I, Robot, Asimov postulated that by 1998, robots would walk well enough to be a child's playmate, but a "Talking Robot" would be an unwieldy thing confined to a museum. Reality has, in fact, given us the opposite—you probably have a "talking" Q&A bot in your pocket, but science has yet to master the biomechanics of human movement.
According to Doctor Singer, legs are an engineering challenge because they combine the tasks of movement and balance. A single step requires dozens of muscles, joints, and ligaments working in concert from toe to hip, with the brain and body making instinctive adjustments on the fly. Robots aren't quite there yet. This complexity means walking robots—even the most famous versions like box-carrying Atlas at Boston Dynamics—are in what Singer describes as the "science project" phase.
"There's work going on but not a single one of them has been deployed," says Singer. "Not one of them has been shown in working prototype form at a trade show. We're not even at that point."
While anthropomorphic robots look promising in the lab, he says, right now they're mostly good for making cool YouTube videos. Many can't function for long without being tethered to external power sources. On the other hand, other kinds of military robots—ones that move on wheels, tracks, wings, and propellers—are already seeing use on the battlefield. "Legged is still in science project, the other forms, we can go touch them, we can see them operate, we can buy them."
Above: Boston Dynamics' "Spot" robot.
When Singer advises both game developers and military scenario-writers, he likes to keep them realistic about the pace of robot development. That's why the near-future set Black Ops II—which Singer consulted on—has insect drones, autonomous aircraft, and miniature tank sentries, but no humanoid military robots.
"If it's ten years in the future, it's hard to think you're going to have lots of legged systems," he says. "If you're 50 years out, then these issues will be solved."
But we will eventually see legged robots, he says—because these systems are highly desirable for urban warfare.
Human Bodies to Navigate Human Cities
Part of the fixation on humanoid robots is a failure of imagination, a clinging to what we've seen on screen. But in the future, anthropomorphic robots will almost definitely deploy to the battlefield. Given the rate of urbanization and population growth in coastal cities, the Army War College predicts that the coming decades will see more battles in heavily urbanized environments—in other words, spaces specifically built for things that look and move like humans.
"If you're thinking about a robotic system to move through a building," says Singer, "Guess what? We designed the building for people of a certain height or shape. It's a very real design problem. We built the world a certain way."
In fact, researchers are currently trying to build robots that can navigate basic human architecture. The 2008 TechX Challenge in Singapore offered a $700,000 prize to any team whose robot could autonomously enter a building, go up some stairs, and press an elevator button for a specified floor. No one won. Most robots didn't even get in the front door—which they didn't even have to open.
However, the tech is advancing. Recently, a team at Stanford built a robot that could not only push the correct floor with regularity, but recognize the correct button on elevator panels it had never seen before. It was a huge leap forward. This robot had both functioning eyes and a "finger"—two crucial pieces of equipment humans need to navigate buildings. But stairs—a human invention just as "unnatural" as an elevator—could stop it dead.
By contrast, Explosive Ordinance Disposal (EOD) bots currently in operation generally work on wheels, tank-style tracks, or both. Many can tackle stairs, if a bit clumsily.
But according to Singer, legged systems are still a prize worth pursuing. Different movement systems are better for different types of terrain—wheels are fast but need flat surfaces, tracks sacrifice speed for the ability to mount rubble—but human legs are versatile. "You can walk in anything," he says, "perfectly flat terrain, up stairs, or up the side of a mountain. You can walk on concrete but also on sand."
Navigating built space is especially desirable, though, since the military is interested in developing robots that can clear buildings—a notoriously dangerous procedure. One bot currently serving this role is the 110 FirstLook by Endeavor Robotics, a man-portable tracked drone that can be thrown through a window and remotely controlled, letting soldiers explore the building via the bot's built-in camera. Future manned and unmanned systems, the Pentagon hopes, will range from "Iron Man" exoskeleton suits to door-breaching bots that let soldiers enter and clear rooms without risking casualties. These systems could protect soldiers' lives, but make war even nastier for anyone who faces them.
Both Titanfall and Infinite Warfare display this dynamic. Infinite Warfare's shield robots fill this door-breaching role for the Settlement Defense Front, chasing the player out of their position and frequently storming rooms. Meanwhile, Titanfall 2 lets the player "hack" robots into allies, forcing them to take bullets for you as they soften up the enemy. Both scenarios are an acknowledgement that robots may one day do the jobs that prove too lethal for flesh-and-blood soldiers—a pattern that's developing as we speak.
A Robot for Every Job, and a Job for Every Robot
The robots in Titanfall 2 and Infinite Warfare aren't just combat models though—they also do industrial work. Crawl onto the skin of a starship or delve into a mine on Mercury and you're sure to find robots doing jobs in conditions where humans wouldn't survive.
Which is, after all, how we use military and industrial robots today—unlike their UAV cousins in the sky, our current ground-based robots are mostly focused on preserving human life rather than ending it.
Current military robots mostly serve Explosive Ordinance Disposal (EOD) functions, or operate in the presence of radioactive or chemical material. After the Fukushima Daiichi meltdown in 2011, for example, Tokyo Electric Power Company hired a contractor to survey the site via a T-Hawk drone. This use of drones at Fukushima continues today, and last year, a purpose-built drone managed to broadcast video and radiation readings from inside the reactor itself, surviving for three hours under conditions that would kill a human in seconds. This is, in some ways, a best-case scenario for robotic systems—taking over tasks that in earlier meltdowns, like Chernobyl, killed the response workers trying to help.
As technology advances, defense planners are looking at other dangerous jobs robots might replace. Self-driving convoys, for instance, could help mitigate the number of personnel lost to IEDs. Walking or wheeled robots could help increase the amount of supplies teams can carry in the field, cutting down on the need to return to base or expose troops to fire while ferrying extra ammunition or equipment to a firefight.
Earlier this year, Israeli company Tactical Robotics conducted the first successful test flight for the AirMule, an autonomous air ambulance drone meant to land and take off in battlefield conditions too space-restricted for traditional helicopter evacuations. Johns Hopkins recently proved you could deliver blood via drone, and Japan has begun testing them for defibrillator-delivery.
You might notice that these drones, overall, look more like the equipment they're replacing rather than the people whose jobs they take over—and that's by design. A robot's task is to fulfill its function, after all, not to fit mimic humans. A robot medic is less likely to look like a human medic, and more likely to resemble the proposed Robotic Extraction Vehicle/Robotic Evacuation Vehicle (REX/REV): a self-driving ambulance that deploys a stretcher bot optimized for carrying and treating casualties. The reasons for this are largely practical, says Dr. Singer, and driven by market forces.
"If a factory is set up to make cars, the conversion to making a robotic car saves money if you keep doing the same thing," he says. "We're, in essence, changing the software rather than the hardware." He adds that customers—including the military—are also more comfortable buying equipment that looks familiar.
This is especially important in the military, where the generals who have final say on acquisitions are usually older and less comfortable with tech. For example, last year a captain at the U.S. Army Cyber Institute built a drone-jamming machine with a Raspberry Pi—it could've looked like a radio, but he made it look like a gun so top brass would understand what it did. "By putting all of this stuff on the rifle frame, it also makes it very easy for senior leaders to consume," he told Popular Mechanics. "Aim. Shoot. Crash."
That ability to describe complex technology to higher-ups will be key in the next administration. In a 2007 deposition, president-elect Trump admitted that he did not own a computer and he appears to read emails and online articles in hard copy. To officials like Trump, familiar-looking technology may have an edge in purchasing decisions.
The problem with simply modifying current vehicles, however, is that it holds back innovation and efficiency. Mine-clearing robots that look like tractors may sell better for now, but at some point, that empty driver's cab will be considered wasted material and eliminated. In fact, part of the promise of autonomous systems is that they can eliminate unnecessary systems, saving fuel and creating more space for cargo, payloads, or equipment.
Similarly, robot technology may be lagging behind due to its effort to replicate human locomotion rather than improve on it. While it's true that legged movement is an excellent traversal system, bipedal legs aren't necessarily the most efficient version. Multi-legged robots could provide a more stable firing platform and solve balance issues in the short term—meaning the first functional legged robots might resemble dogs or spiders rather than people. Incredibly, future war robots might take their design cues from nature, like those in upcoming caveman vs. robots game Horizon: Zero Dawn.
Above: the Horizon Zero Dawn trailer from PSX 2016.
There are already plenty of projects attempting to replicate biological systems in robotic form. European researchers are working on an artificial eye that mimics the curved compound eyes of an insect. MIT produced a cheetah-inspired robot—with funding from DARPA—that can run at high speeds and leap over obstacles. Other scientists are pursuing drone systems that sip energy off power lines while perching on them like a bird, or "eat" organic matter and burn it for fuel (that robot fell out of favor when journalists discovered it could, theoretically, live off human corpses). One team of researchers want to build a sensor that replicates the hawk's unique vision, which is a bit like the picture-in-a-picture setting on TVs.
"Hawks have the ability to scan broad areas but zoom down on a target simultaneously," says Singer. "That's something we're looking to equip our high-altitude drones with."
But the most interesting thing robots are borrowing from nature, he says, is disaggregated intelligence. We tend to think of robots as a large intelligent system, like humans, but there's an increasing amount of interest in swarms of tiny robots that each have a specific task, much like ants, and can coordinate for attack or surveillance missions.
"Each individual ant is not all that smart," says Singer, "but working together they can do highly intelligent tasks. We're seeing the same split with regard to robotics—are you going to do the task through one big super-smart system, or are you going to do it through lots of little tiny parts?"
Drone swarms aren't speculative technology, they're already with us. Disney just launched an aerial light show in Orlando where 300 LED-lit drones form pictures of Christmas Trees and flying birds, with the programmed flight choreographed to music. It's not hard to imagine a terrifying future where, instead of LEDs, those drones carry tiny cluster bomblets similar to the ones Hezbollah drones have dropped on Syrian rebels.
With some researchers studying ants, and others trying to replicate hawks, it's possible that future robots—even if they look like humans—may actually be technological hybrids of many different creatures. Future war robots might walk like cheetahs, see the world like a fly, and have arms like a human.
And since robots are not natural creatures, there's no need for them to obey biological or evolutionary laws. There's no animal on Earth that has eyes in its hands—but modern EOD bots already have this feature, since it assists the delicate process of disassembling IEDs. Rather than humanoid or animalistic, war robots may end up looking like something out of an H.P. Lovecraft story, bristling with arms and eyes.
That's intriguing to think about, until you imagine yourself as the poor flesh-and-blood soldier or civilian who has to face that monstrosity on the battlefield.
Fiction Follows Reality, and then Reality Follows Fiction
But despite these possibilities, the game industry mostly insists on sticking with the traditional sci-fi concept of military robots—one shaped more by the Tin Man and Metropolis than the latest in real-world robotics. As an advisor to game studios, Singer tries to make the case for down-to-earth robot systems, but artists often cling to preconceived notions.
"You'll be like: 'That's not realistic,' and people will say: 'But that's how robots are supposed to look,'" Singer explains. "Science fiction shapes the way we think of the future."
But when you keep feeding them into the culture, he says, these depictions of robots create a feedback loop, where military leaders see Hollywood-style robots and believe that's where the technology will inevitably lead. "It's the same problem on the military side: thinking that this [design] must be futuristic because it looks futuristic—it's a virtuous cycle."
But even though Titanfall 2 and Infinite Warfare lean heavily on robot tropes, they do get a lot right. According to current defense theories, robots will accompany and augment traditional military units rather than replace them. The Marine Corps Warfighting Laboratory envisions the armed MAARS system—essentially a remote-controlled tank mounted with a camera array, an M240 machine gun, and a grenade launcher—or the larger RV-M as an extra squad member who can carry heavy weapons and perform limited recon.
Meanwhile, Russia's Uran-9 unmanned tank hopes to pack both anti-infantry and anti-armor lethality into a small package, with the crew operating the machine remotely at a safe distance. Though these miniature tanks look nothing like the robotic squadmates in Titanfall 2 or Infinite Warfare, their function is essentially the same—augment troops' firepower while keeping them safe.
When Reyes commands an armored mech in Infinite Warfare, he essentially operates the system remotely, much like the Marines currently envision for MAARS or the Russians the Uran-9. Likewise, the Seeker grenades in Infinite Warfare—crab-legged explosives that find enemies behind cover—are just an autonomous version of what Dallas Police did this summer, when they strapped C4 to an EOD robot and ran it into a sniper holed up behind cover. That alarming incident—the first confirmed case of an unmanned ground vehicle (UGV) killing a human—prompted a call for police to re-draft their use-of-force regulations and provide clear guidelines for when it's appropriate to use a lethal drone.
The good news is that the Department of Defense, perhaps learning from cavalier use of UAVs in the War on Terror, have been slow to introduce Unmanned Ground Vehicles. The predecessor to the MAARS system, SWORDS (which is also operated via remote control) deployed to Iraq but never fired a shot at a hostile target, partially due to minor targeting issues discovered during testing. The Army Program Manager responsible suggested that the Pentagon was merely being careful, believing that if a UGV either underperforms or makes a targeting error early on, it would set the program back decades.
But the Pentagon's kid-gloves attitude toward unmanned systems may shift dramatically in the coming months, when a new administration can rewrite the rules involving robots in war.
The Trump Administration Will Decide the Future of Robotic Policy
According to Singer, the rise in automation will have major consequences legally and economically, and the Trump administration will decide the future of robots in both war and peace. Self-driving cars will mean accidents with no person clearly at fault (was it the sensors that failed, or the breaks?) but it will also lead to job losses in areas like transportation and manufacturing.
"It's not only a legal question of how we handle robotic systems in war and on the highways, it will continue to drive this sense of economic displacement people have," Singer says. "It's playing out in China too, with the Foxconn factory—they're increasingly being automated. A few weeks ago, CNN had one of those isn't-it-neat stories about a robot that makes an entire shirt on its own. What does that mean to Bangladesh?"
But apart from the economic consequences of automation, the Trump administration will have to make a determination about U.S. military policy toward Lethal Autonomous Weapons, or LAWs. Current Pentagon policy, set in 2012, stipulated that there always needs to be a human in control of whether autonomous systems fire or not—the weapons cannot kill on their own. The guidelines are vague, and can be waived by senior leadership if deemed necessary, but stipulate that a human should make the call.
"It had a clause that said this policy had to be renewed or not in five years," says Singer. "Next year's the five-year point."
In other words, the Trump White House will decide if killer robots become a part of America's arsenal—or whether the decision between life and death remains in human hands.
Robert Rath is a freelance writer, novelist, and researcher based in Hong Kong. His articles have appeared in Zam, Vice, The Escapist, Playboy and Slate. You can follow his exploits at RobWritesPulp.com or on Twitter at @RobWritesPulp