When the Killer Robots Arrive, They'll Get Hacked

But it may be awhile before those systems are even developed.

Feb 24 2015, 5:10pm

​Engineering students from Penn and Virginia Tech demo the Shipboard Autonomous Firefighting Robot. Image: ​U.S. Naval Research Laboratory/Jamie Hartman

Autonomous weapons could be hacked and turned against us, said Peter W. Singer, a strategist and senior fellow at the think tank New America Foundation, at the first annual Future of War Conference on Tuesday. We are entering "a whole new realm" when it comes to autonomous weapons, he argued as part of a panel discussing the future of smart weapons.

"Cyberwarfare offers you the ability to persuade the target to do something it wouldn't do otherwise," he said. "I gain access to its software, if I [hack] into it, I can then make it do things other than the operator wants it to do."

He said this includes hacks as simple as changing the GPS location of a drone, which he said has been done with relative ease by the Department of Homeland Security and by college students, or what he called "ultimate co-option," which would be, for example, "recoding all American systems as Chinese systems."

"A human would think, 'That makes no sense, I'm questioning that order,' but a computer, if you have that access, will follow that instruction," he said. "It's a whole new realm where you've never been able to convince a bullet or an arrow to change direction in mid-flight, you can with this kind of system. That points to both possibilities and perils."

We definitely do not have the capabilities to have a robot able to reason in all kinds of areas of uncertainty

However, it's unclear how long it will be until we see technology like this come to fruition. Missy Cummings, director of the Humans and Automation Laboratory at MIT, said "we're not even close" to developing autonomous killing machines in the sense that many people think of them.

"We are simply not in a place where the technology has gotten to the point we are going to have have truly killer weapons," she said. "Humanoid robots are not going to become a reality on the battlefield for the foreseeable future because of energy problems. We simply cannot build one without a battery the size of a tank."

The artificial intelligence technology is also not up to speed for autonomous weapons to be reaching battlefields anytime soon, she argued.

"For example, with drones, we have to have the operator make the decision, because the artificial intelligence is not up to that level," she said. "We are still in its infancy. We definitely do not have the capabilities to have a robot able to reason in all kinds of areas of uncertainty."

She also said the US is trailing behind several other countries in this realm of technology.

"My students can, over the weekend, build a drone more capable than what many US military personnel have access to," she said.

Although these experts say we are a long way from seeing autonomous cyborgs on the battlefield, it's clear the government is actively developing these technologies, bringing up a variety of social and political implications.

Activist groups are already campaigning against "killer robots," and in April, world leaders will meet in Geneva to discuss their regulation at the Convention on Certain Conventional Weapons (CCW).