Is This a Good Idea? Killer Robots?

July 01, 2009

    Should autonomous robots —that is,  robots who can perform tasks in unstructured environments without continuous human guidance--be armed with lethal weapons and allowed to decide for themselves whether to kill humans?

        To be sure, Killer robots would have plenty of useful applications.  Law enforcement versions could protect banks from robbers, patrol dangerous neighborhoods and even keep guard over convicts in prisons without any actual human officers being endangered.  In the military, they could take the place of human soldiers, venturing into battlefields and enemy territory on missions too hazardous for real troops. If they get blown up by an IED, they could be repaired or replaced, without the need for a hospital stay and lengthy, painful rehabilitation. An army of killer robots wouldn’t run the risk of developing Post Traumatic Stress Disorder and they wouldn’t become despondent about a spouse and kids back home who miss them. 

On the other hand, if you’ve seen Terminator Salvation,  you can easily imagine the downside of autonomous armed robots. What if, instead of dutifully protecting us good humans from bad ones, they choose to follow the old martial credo of “kill ‘em all and let God sort ‘em out” and turn on us all?

We’re in potentially even worse shape if those robots have the ability not just to kill but also to create additional, even more advanced versions of themselves. (While that capability may seem like a sci-fi fantasy, the Pentagon actually is interested in developing something called a Self-Explanation Learning Framework (SELF) —basically, machines that can reason, analyze their own behavior and even  “participate in their own construction.”)

But even if they don’t decide to wage a genocidal war against their former masters, killer robots pose plenty of other potential problems. If human soldiers often have difficulty distinguishing between enemy combatants and innocent civilians, how well could machines tell between the two? And what if the technology is obtained by bloodthirsty dictators, criminals or terrorists, who figure out how to override whatever ethical restraints are built into the robots’ software and turn them into remorseless, pitiless murder machines? 

        The idea of killer robots has been around at least since
Karel Čapek’s 1920 play R.U.R. (Rossum's Universal Robots) the work of literature that popularized the term “robot.” In the drama, a race of worker androids revolt and exterminate their human masters, save  for a lowly clerk named Alquist.  In the 1940s, science fiction author Isaac Asimov solved the problem of future robot malevolence in the future with his First Law of Robotics which stipulated that

        A robot may not injure a human being, or, through inaction, allow a human being to come to harm.

        Asimov’s sci-fi contemporary Philip K. Dick, in contrast, in his 1953 story “Second Variety,” envisioned the aftermath of a World War III in which U.S. battlefield robots, after wiping out the Soviets, created and deployed next-generation android assassins masquerading as surrendering survivors to finish off the remaining Americans in bunkers.  (Dick’s story bears unmistakable similarities to the premise of the later Terminator movies.) 

t’s only been in recent years, however, that robotics and artificial intelligence capabilities have begun to catch up to sci-fi authors’ imaginations.  In the Iraq war, the U.S. military has utilized thousands of battlefield robots such as QinetiQ’s TALON, a remote-controlled vehicle that has helped soldiers clear IEDs and perform surveillance. An armed version of the TALON, called SWORD, which can be outfitted with a machine gun, a grenade launcher and other weapons, also has been deployed, but apparently has yet to see actual combat.

National Defense magazine reported last May that the first three SWORDs had been stationed behind sandbags rather than sent to patrol Iraqi streets, after an incident in which a robot’s turret moved without any human instructions. Additionally, the magazine noted,

Detractors have questioned their vulnerability, claiming an enemy soldier could defeat it by sneaking up from behind with a baseball bat or by tossing a blanket over it. The first generation SWORDS cannot swivel around 360 degrees.

Already, the same manufacturer has developed a more advanced armed robot, MAARS, that it touts as having a “transformer-like” ability to change shape, akin to the Hasbro action figures and the movies inspired by them. 

Both SWORD and MAARS are remote-controlled, South Korea already is using a stationary armed robot sentry that’s capable of operating autonomously along its border with North Korea, and my guess would be that it’s only a matter of time before the Pentagon develops an autonomous armed robot, capable of negotiating the battlefield—and engaging the enemy—on its own.

John Pike, a defense and intelligence expert and the director of GlobalSecurity.org, has predicted that such systems will be operating by 2020. 

Maybe that won’t be such a bad thing. Georgia Tech robotics researcher Ronald Arkin, author of the just-published book Governing Lethal Behavior in Autonomous Robots, argues that self-controlled killing machines might actually lead to more  humane warfare, because they would be better at avoiding the inflicting of civilian casualties than human soldiers, whose actions are influenced by emotion and the fog of war. In this paper on the subject, describes how robot behavior might be designed to incorporate ethical restraints.

But others fear the worst. Noel Sharkey, a professor of artificial intelligence and robotics at the University of Sheffield in the UK, doubts that killer robots can be programmed to discriminate effectively. In a 2007 column for the Guardian, a UK newspaper, he argued that

In reality, a robot could not pinpoint a weapon without pinpointing the person using it or even discriminate between weapons and non-weapons. I can imagine a little girl being zapped because she points her ice cream at a robot to share. Or a robot could be tricked into killing innocent civilians…machines could not discriminate reliably between buses carrying enemy soldiers or schoolchildren, let alone be ethical. It smells like a move to delegate the responsibility for fatal errors on to non-sentient weapons. 

Human soldiers have legal protocols such as the Geneva conventions to guide them. Autonomous robots are only covered by the laws of armed conflict that deal with standard weapons. But autonomous robots are not like other weapons. We are going to give decisions on human fatality to machines that are not bright enough to be called stupid. With prices falling and technology becoming easier, we may soon see a robot arms race that will be difficult to stop.

In a March 2009 McClatchy Newspapers article Sharkey put it even more gravely:

We are sleepwalking into a brave new world where robots decide who, where and when to kill.

So what do you think? Hurry up and express your opinion below, before the killer robots come for you.

Comments

[Maren Wryn]

Fascinating premise, several actually. The problem with the development of autonomous robots, especially the kind that develop a further understanding of their environment on their own, is that they are designed to do so by humans and are, therefore, destined to acquire all of our human flaws. The Asimovbot just won't ever exist if they have any autonomy at all. We have killer robots now, in the form of unmanned drones flying over Iran and Afganistan. To be sure, a human operator is simply using a remotely controlled flying object as a platform, but the step to autonomy from that platform is very very low. So low, in fact, it could be a tripping hazard. Next step - the drones "hunt" for preprogrammed targets and "request" permission to kill. Once that's worked out and their requests are about 90% appropriate - it's a short hop to "Go for it" and launch. It's coming whether we like it or not. Good of you to have pointed that out. Thanks. @OregonMJW

[nathan jones]

Even though using robots will save the lives of alot of soldiers there is still absolutley no technology in the world can replace a soldiers keen instinct on whether or not to kill a human and besides, what if the robot malfunctions and ends up killing innocent people if robots are given the ability to make there own descisions then who knows what they'll do!?!

[Mothra]

Actually, I think it's a really terrible idea. Who programs the robots, and what are their MOTIVES?

[Minh]

IT PRETTY GOOD BUT IF SOME ONE BAD GOT IT IT WOOD BE BAD

[Caffeine Driven Stress Magnet]

This is a scary idea, but it will probably happen soon.

[Abouti]

If the Iranian government had killer robots, it would have been even easier for them to put down the protests over the election.

[Astroboy]

HAPPY 4TH OF JULY EVERYBODY!!!!!!

[Patriot]

I think an armed robot would be a great way to protect your house from intruders. The NRA has to get on this right away, to make sure that the government doesn't outlaw the private ownership of armed robots.

[Dr. Strange]

what protection would we have against killer robots running amok?

[Speed Racer]

What if hackers got control of the robots?

[Bryan]

I think with robots that we should see them as children, and what I mean about that is that when we firts make robots we should not be showing them violence and destruction but as there main purpose but what you would show a first grader. Then once we are more aware of robots and seen its advantages and weaknesses then we can start making robots not to kill but to just capture or bring down the criminal or suspect. Seriousele what are people thinking of making a machine that can kill humans without any truoble haven't you sciencetist seen the movies?

[armani webb]

I dont like that idea, what if they see humans as threats a few years after there creation. were done for it if that happens. It's like commiting suicide but to the whole human race.

[John Connor]

Take my advice--this is a bad idea. Put a stop to it now, before it's too late.

[Dandroid]

I don't know why people are so upset about the idea of robots having the power to kill humans. It's not like they don't deserve it.

[Ian]

It's time to just turn over control of everything to the machines. Just kidding. This is a terrible idea.

[Daniel Jones]

I'm a pacifist so killing machines, I vote NAY.
but I'm 13, what do I know?