AI-powered armed robots trigger a debate around safety, future of warfare

US, Israel, Korean Demilitarised Zone, Gaza, all have these robots in one form or the other

The San Francisco Police Department’s (SFPD’s) decision to induct robots that can use deadly force has led to controversy in America’s civil society and upped decibel levels in an ongoing debate about “killer robots” – that is, semi-autonomous and fully autonomous weapons. The SFPD could deploy robots “equipped with explosive charges to breach fortified structures containing violent, armed, or dangerous subjects".

The Dallas police have already used a robot, the Northrop Grumman Remotec, with a payload of explosives to winkle out an urban sniper hidden inside a building. Apart from explosives, the Remotec is marketed as capable of using a shotgun, a gas dispenser for crowd control, a window-breaking tool, cable-cutters, drills, saws and such. It can also deploy several different rocket launchers and rocket-propelled grenades (RPGs).  

That’s lethal enough. But military drones carry a great deal more in the way of munitions and of surveillance equipment. Military robots come in many shapes, sizes and use-cases, ranging from aerial and marine drones to land vehicles.

As the Ukraine war demonstrates, cheap drones can be devastatingly effective even against sophisticated defensive systems. The US has also used (exceedingly expensive) drones like the Reaper (the MQ-9) to loiter in the sky for over 24 hours at a time, to find and identify specific targets, or to even target individuals.

Defence analyst Pravin Sawhney, in his recent book The Last War: How AI Will Shape India's Final Showdown With China, outlined a credible scenario where the Chinese use aerial drones, hovercraft and other killer robots as frontline “troops” in a 2024 invasion across the Himalayas.

Such units all run on artificial intelligence (AI), which manages their operations and makes them nearly autonomous. They can be remotely controlled by operators who “pull the trigger”, or give them specific instructions in terms of surveillance tasks, for example.

It doesn’t take much imagination to realise the next generation of such robotic weapons systems could be fully autonomous, or closer to it. The major difference is that there wouldn’t be a human finger on the trigger – the robot would make its own decisions on when to fire and at what.

It would not be technically too difficult to move to folly autonomous drones.

Fully autonomous drones could receive broad instructions to find and kill a specific individual, or fire missiles at specific targets, or attack targets of opportunity, such as tanks, ships or truck convoys.

Intelligent drones with embedded face recognition technology could be used for targeted surveillance, or assassination. Smart machine guns with face recognition and voice recognition could defend installations against anybody who wasn’t in the audio-visual database.

Defence and offence

Autonomous defensive systems already exist. Not having humans in the loop improves response time, which is critical in many cases. 

There are sentry guns deployed on the edge of the Korean Demilitarised Zone and near Gaza. There are autonomous anti-missile systems such as Israel’s Iron Dome, and multiple types of anti-aircraft systems.  

The ethics of creating AI and deploying it in such ways is one of the hot-button issues in this debate. Making more killer robots fully autonomous is another related issue.

While plenty of people have voiced unease at the prospects, every military-industrial complex worth the name is also pouring resources into R&D leading in these directions.

“Killer robots” would eventually become “off-the-shelf” technology used by military, law enforcement, terrorist and criminals alike. That’s inevitable.

The beauty (and danger) of AI is that smart teenagers with normal access to consumer electronics can cobble together smart devices.

Houthi rebels have used commercial drones with attached IEDs (improvised explosive devices) to attack Saudi installations, for instance. The military tech is only a little more sophisticated than the commercially available stuff.

As technology advances, “amateurs” will put together ever more lethal variations. The classic crazed American shooter could be supplemented by crazed individuals deploying swarm killer robots.

There have been attempts in the tech community to move away from militarisation.

In October for instance, six robotics companies pledged not to weaponise in an open letter which said in part, “We believe adding weapons to robots that are remotely or autonomously operated, widely available to the public … raises new risks of harm and serious ethical issues.”

There have been movements within Google, Amazon and Microsoft where employees (who are also shareholders, of course, in these instances) have protested the digital giants’ picking up defence-related work contracts.  

However, unlike nuclear research, which requires huge government resources and cannot be done in a school lab, AI R&D can be done anywhere, by pretty much anybody. Given what’s at stake, money will continue to flow into this area. While it’s true that the benefits from AI could outweigh the risk of misuse, killer robots could soon become a common hazard.