Lethal autonomous weapons (or killer robots as the media likes to call them) was the subject of intense discussion in the corridors and committee rooms of the United Nations in Geneva. The international talking shop played host to the third round of multilateral talks on this topic. The meeting followed on from increasing concerns about the rapid progress being made in areas like artificial intelligence (AI) and robotics. Stephen Hawking, Elon Musk, Bill Gates and others have expressed concern about the direction these technologies may be taking us.
Military robot image from Shutterstock
Toby Walsh is the Professor of AI at UNSW and Research Group Leader Data61
Last July, thousands of researchers working in AI and robotics came together and issued an open letter calling upon the UN to put a pre-emptive ban in place on such weapons.
In the interests of disclosure, I helped put the letter together and will be talking at the UN meeting on Thursday.
Where will this end?
If we don’t get a ban in place, the end point is clear to my colleagues and me: there will be an arms race and it will look much like the dystopian future painted by Hollywood movies like the Terminator series.
The technology will undoubtably fall into the hands of terrorists and rogue nations. These people will have no qualms about removing any safeguards in place on its use. Or using it against us.
Unfortunately, we won’t simply have robots fight robots. Wars today are asymmetric and it will be robots against humans. Any many of those humans will be innocent civilians.
This is a terrifying prospect.
We don’t need to end there
The world has come together in the past to decide not to weaponise a technology. We have bans on biological and chemical weapons. We have treaties to prevent the proliferation of nuclear weapons.
Most recently, we have collectively agreed to ban several technologies including blinding lasers and anti-personnel mines.
And whilst these bans have not been 100% effective, the world is undoubtedly a better place for their existence.
The treaties have also not prevented related technologies from being developed; you go into a hospital, and a “blinding” laser will be used to fix your eyes. But if you go to the battlefields of the world today, you will not find blinding lasers being used. And no arms company today will sell you one.
The same is likely to be true for autonomous weapons. We won’t stop the development of the broad technology. It’s much the same that will go into an autonomous car as an autonomous drone or submarine.
And we’ll definitely want autonomous cars. One thousand people will die on the roads of Australia this year. These numbers will plummet once we have autonomous cars. Most accidents are the result of driver error.
But if we get an UN ban in place, we’ll not have autonomous weapons out in the battlefield. And this will be a good thing.
Image: The General Atomics MQ-9 Reaper is already semi-autonomous, and similar combat aircraft could soon be fully autonomous/USAF Photographic Archives
Come on Australia
Australia has led the world in many discussions around disarmament. For instance, we have taken a leading role in nuclear non-proliferation.
But we have taken a disappointing role so far in the UN discussions around autonomous weapons. Our official position appears welcoming.
The development of fully autonomous systems able to conduct military targeting operations which kill and injure combatants or civilians may be closer than many of us had imagined. It is an appropriate time to consider the risks of such weapon systems and to make sure we understand fully what might constitute misuse as well as legitimate use of emerging technologies.
However, we are not helping the discussion with official statements like the following.
If we were to settle, ultimately, on an agreement that there were limits to the autonomy that lethal weapons may possess, or that there were limits to the weaponisation of autonomous systems, we would also have to design ways, not just of defining, but of implementing, such limits, and of verifying compliance. We should not underestimate the complexity of this task.
This is not just unhelpful but also wrong. There is no necessity to define ways to verify compliance. Almost no weapon banned by the UN has a compliance regime.
There is no international body to inspect for blinding lasers. Or anti-personnel mines. Even the grand-daddy of all weapon bans, the 1975 UN convention on biological weapons, has no formal compliance measures beyond self-reporting by nation states and investigation by the UN Security Council (which has never occurred).
There is also no necessity to define limits on autonomy. For example, the 1998 UN Protocol on Blinding Laser Weapon does not formally define a limit on the wavelength or wattage of a “blinding” laser.
We can simply require that autonomous or semi-autonomous weapons must have “meaningful” human control. And depend on the consensus that will undoubtably emerge internationally as to what precisely this means.
Let’s take the lead
Australia is a world super power in AI and robotics. We punch well above our weight. We have some of the most automated ports and mines in the world. And we are currently reigning world champions at robot soccer. Indeed, we have been world champions, so far, five times.
And from the reaction I have had talking about this issue in public, the general population here in Australia supports the view held by both me and thousands of my colleagues that a ban would be a good idea.
All technology can be used for good or bad. Australia should be taking a lead in pushing the world down a good path.
This article was originally published on The Conversation.
Comments
11 responses to “Opinion: Australia Should Take A Stand Against ‘Killer Robots’”
There are a few issues with this idea of banning something, what do you do when someone does not agree with you? are you going to go to war with China? Put trade sanctions of them when they are one of the worlds biggest trading partners?
If you do build killer robots you are also able to build and test counter measures against them, if you do not build killer robots then how would you know your counter measures would work when someone does build a killer robot.
Researching in robot development will not stop and it will not take a AI Genius to add weapons to an existing robot.
If everyone has killer robots then the only thing killer robots will be killing is other killer robots that are trying to kill them.
Is there any actual evidence that remotely controlled robotic weapons are more dangerous to civilians than human soldiers with a personal presence are?
This claim seems *very* counterintuitive to me.
If you’re only there as a robot, surely there’s no fear-induced impairment of judgement.
An autonomous weapon with tracking and firing skills that are hundreds of times faster than humans could do phenomenal amounts of damage.
Good!
Because they are fast and tough and fearless, they can take the time to dispassionately identify and exclude non-combatants better than human warriors!
Depends on (1) whether they are programmed to do such, and (2) they are able to do such.
I refer interested people to Philip K Dick’s short story “Second Variety”
(1) seems irrelevant, we can choose how we program robots, we cannot choose how others do so. Pretending guns don’t exist is also a failed strategy.
(2) Fails on two counts.
2A: They only need to be better than humans at this evaluation, and lacking fear, having more senses, and posessing superior speed will all help.
2B: Sufficiently superior combat robots can be made better than humans at non-lethal and non-harmful takedowns. “I’m terribly sorry our robot had to sedate you sir, we only did so to safely reach the gentleman who was using your son as a human shield.”
I acknowledge the existence of guns and the capacity of well-programmed robots which I was I was commenting on your original question/assertion.
I’m more worried about domestic application of such technology as we’re at the point where autonomous weapons make lone gunmen as effective as a platoon, moreso if harnassed to flight capacity. Peter Singer (strategist, not philosopher) is on top of this stuff as usual.
Australia needs to make ALL OF THE KILLER ROBOTS!!!
Base them on Aussie animals.
Anyone attacking us will live in utter fear of ambush by Drop Bear or having to face the devastating ComBat Wombat tanks as they roll over everything in their path.
Didn’t you know australia was an abandoned extraterrestrial bioweapons facility?
And we need to capitalise on that technology, dammit!