15 June 2009
March of the killer robots
The development of mechanical soldiers and remote-controlled tanks and planes is changing war for ever - but the moral consequences have often been overlooked.
By Thomas Harding
Daily Telegraph Defence Correspondent,
It's the most realistic shoot-'em-up game ever. The player has a choice of two planes: a Predator with two Hellfire missiles, or a Reaper with 14. The action takes place in the Middle East, where you can attack villages and kill the inhabitants with impunity. But don't bother looking for it in the shops: to play this deadly game, you'll have to travel to Creech Air Force base in the Nevada desert. That's because the planes are real, and so are the casualties.
The first time a Predator made a kill was in Yemen, in 2002, when the CIA used it to destroy a vehicle carrying an al-Qaeda leader and five of his associates. The fleet now stands at around 200 craft, which have flown more than 400,000 combat hours. The company that makes them, General Atomics, can't keep up with the demand. The bigger, badder version – the Reaper hunter-killer – is also flying off the shelves. There are now around 30 in active service, with the first kill taking place in the mountains of Afghanistan in October 2007.
In every field of warfare, mechanical soldiers are fighting alongside – or instead of – human beings. Apart from unmanned combat air vehicles such as Predators, the skies above Iraq, Afghanistan and Pakistan are filled with drones carrying out surveillance operations. On the ground are between 6,000 and 12,000 robots, up from a mere 150 in 2004. Their role is mostly to protect our soldiers by disrupting improvised explosive devices, or to carry out surveillance of dangerous places such as caves and buildings.
Our image of such robots owes a great deal to films – most notably The Terminator or Transformers, both of which have sequels out this month. But the actual models being used are more like miniature tanks, similar to the contraptions seen on the television series Robot Wars. The most popular is the PackBot made by the US company iRobot, which is normally used for bomb disposal. As the company started out making robotic vacuum cleaners known as Roombas, the 18kg PackBot is sometimes jokingly referred to as the "Roomba of doom" or "Doomba" – much to the displeasure of the firm's management, who would clearly hope to keep the two brands separate.
Recently, iRobot joined forces with Taser International to mount the allegedly non-lethal weapons on the "bots". But that pales in comparison with the ordnance that comes with the Talon, a larger device made by Foster-Miller, a US subsidiary of the British firm QinetiQ. It comes with chemical, gas, temperature and radiation sensors and can be mounted with a choice of grenade launcher, machine gun, incendiary weapon or 50-calibre rifle. Its bigger brother, the MAARS robot, ups the stakes with a tanklike turret.
Despite planned cutbacks in spending on conventional weapons, the Obama administration is increasing its budget for robotics: in 2010, the US Air Force will be given $2.13 billion for unmanned technology, including $489.24 million to procure 24 heavily armed Reapers. The US Army plans to spend $2.13 billion on unmanned vehicle technology, including 36 more Predators, while the Navy and Marine Corps will spend $1.05 billion, part of which will go on armed MQ-8B helicopters.
Of course, when the military describes such systems as "unmanned", it is stretching the truth very slightly. At the moment, all the armed robots in the Middle East are remote-controlled by humans – there is a "man in the loop" to control them and to decide when and whether to apply lethal force.
But that makes very little difference to villagers in Waziristan, where there have been repeated Predator strikes since 2006, many of them controlled from Creech Air Force Base, thousands of miles away. According to reports coming out of Pakistan, these have killed 14 al-Qaeda leaders and more than 600 civilians.
Such widespread collateral damage suggests that the human remote-controllers are not doing a very good job of restraining their robotic servants. In fact, the role of the "man in the loop" is becoming vanishingly small, and will disappear. "Our decision power [as controllers] is really only to give a veto," argues Peter Singer, a senior fellow at the Brookings Institution in Washington DC. "And, if we are honest with ourselves, it is a veto power we are often unable or unwilling to exercise because we only have a half-second to react."
As Dyke Weatherington, deputy director of the Pentagon's Unmanned Aerial Systems Task Force, points out: "There's really no way that a system that is remotely controlled can effectively operate in an offensive or defensive air combat environment. The requirement of that is a fully autonomous system."
Sure enough, plans are well under way to develop robots that can locate and destroy targets without human intervention. There are already a number of autonomous ground vehicles, such as the seven-ton "Crusher" developed by DARPA, the US military's research agency. BAE Systems, a British defence contractor, recently reported that it had "completed a flying trial which, for the first time, demonstrated the co-ordinated control of multiple Unmanned Aerial Vehicles autonomously completing a series of tasks". The Israelis are already fielding autonomous radar-killer drones known as Harpy and Harop, and the South Koreans use lethal autonomous systems to defend their border with the North.
Many in the military are enthusiastic about such developments. "They don't get hungry. They're not afraid. They don't forget their orders," says Dr Gordon Johnson, of the Pentagon's Joint Forces Command. "Will they do a better job than humans? Yes."
Dr Johnson insists that "there are no legal prohibitions against robots making life-and-death decisions", adding: "The US military will have these kinds of robots. It's not a question of if, it's a question of when."
The problem, however, is that no autonomous robots or artificial intelligence systems have the necessary capabilities to discriminate between combatants and innocents. Compared with the robots in the Terminator films, they suffer from artificial stupidity. Allowing them to make decisions about who to kill falls foul of the fundamental ethical precepts of the laws of war set up to protect civilians, the sick and wounded, the mentally ill and captives. We are already overreaching the technology and stretching the laws of war.
"Unless we end war, end capitalism and end science, the use of robots will only grow," says Peter Singer. "We are building and using machines with more and more autonomy because they are viewed by militaries as useful for war, and viewed by companies as profitable business." Spending on Unmanned Aerial Vehicles is expected to exceed tens of billions of dollars over the next 10 years, and more than 40 countries – including Russia and China – now have their own programmes.
Amid this robotic arms race, there is a sliver of hope. Professor Ron Arkin, of the Georgia Institute of Technology, believes that humans do not have the time to make rational ethical decisions in the modern battlefield. "There appears to be little alternative," he says, "to the use of more dispassionate autonomous decision-making machinery." He has funding from the US Army for research on how to programme ethical rules into robots to stop them causing excessive collateral damage. But this does not get around the problem of how to discriminate between innocents and combatants – and Arkin admits that the technology to fully support his system may not be available for 25 years.
The problem is that it is not just a matter of developing adequate sensors. In complex wars, complex human reasoning is often needed to decide when it is appropriate to kill. Robots do not feel anger or seek revenge – but they also don't have sympathy, empathy, remorse or shame. Nor can they be held accountable for their actions. In subcontracting our wars to our robotic creations, we are abdicating moral responsibility, too.
Noel Sharkey is Professor of Artificial
Intelligence and Robotics at the University of Sheffield