The AI battlefield
We need global norms to address the rise of machine autonomy in armed conflict
In the Mahabharata, Krishna wielded what would today be called a lethal autonomous weapon: the
Such machines could soon be made for real. On November 13, a United Nations (UN) group of experts in Geneva kicks off the first formal inter-governmental discussion on what machine autonomy means for the laws of armed conflict and the future of international security. I have the honour to chair this group, on behalf of 125 states party to the Convention on Certain Conventional Weapons.
Krishna, being a god, had the wisdom not to deploy his awesome weapon — at least, not directly. He used it to block out the sun, which tricked opposing warriors into dropping their protective shield. Ancient India had rules of war, just as we do: they required fighting to cease at sunset.
Throughout history, the capacity to wield new technologies — from gunpowder to nuclear weapons to long-range missiles — has changed how wars are fought, and the strategic balance between attack and defence maintained.
Shaped by technology
The norms around what is considered acceptable in warfare have also evolved in response to new technologies. Since the 19th century, those norms have been codified in international humanitarian law, which is more or less universally accepted as regulating armed conflict among civilised nations.
Recent advances in artificial intelligence (AI) are throwing up a new challenge to these norms: if the weapon fuses with the wielder, who do they apply to and how? Should such a possibility even be allowed?
Reality might not have yet caught up with popular culture depictions of “killer robots” and “conscious synths” demanding their rightful place in society; indeed, such depictions can be a distraction from the complex challenges that do exist. But many technology leaders are worried about autonomous systems taking life-and-death decisions without “meaningful human supervision or control”. The American tech billionaire Elon Musk and over 100 others recently signed a letter warning that the weaponisation of AI-based technologies risks opening a Pandora’s box.
These are not the only concerns about AI. Technologists and ethicists are also grappling with such questions as legal liability when autonomous vehicles share the streets with pedestrians, predictive analytics subverting due process, and the algorithmic entrenchment of human biases.
Walking a tightrope
But AI applications are already a growing reality in areas such as health, finance and retail. Civilian applications of AI technologies will undoubtedly continue apace. And as has been the experience with other dual-use technologies, AI developed for civilian purposes could be repurposed.
How, then, to deliver on the promise of AI while protecting the hard-won tenets of international humanitarian law and respecting the legitimate security and commercial interests of states and industry? This is the question we will be grappling with this week in Geneva.
Mr. Musk’s letter called on the UN to “find a way to protect us from all these dangers”. Some will query if the UN can succeed. The multilateral system is often derided for its slow pace, its obsession with procedure and its opacity to the wider public. In many areas of technological complexity, alternative governance models have emerged, such as the ‘multi-stakeholder’ approach to Internet governance.
A new approach
However, in an era of diffusion of power and mistrust among the major powers, multilateral inter-governmental forums remain the only way to extend norms across the globe. For bad or for worse, governments still decide matters of war and peace. And the UN still offers a neutral venue to bring different points of view together.
The discussions in Geneva are an opportunity to test a new approach, one we might call ‘distributed technology governance’. This means the multilateral system’s search for durable international norms needs to integrate national regulatory approaches and industry self-regulation.
Each level in this chain of subsidiarity — international humanitarian law, national regulations, and industry self-regulation — needs to move in full cognition of the other two. We need to find ways for them to enjoy their respective sovereignty, while working in unison to deliver what the international community expects.
When Alan Turing, the British scientist who can rightfully be called the father of AI, first speculated on the promise of thinking machines, he pointed out their potential for making us think about ourselves — our faults, frailties and foibles. Aspiring to the wisdom of Krishna may be expecting too much, but we should welcome the fact that AI challenges us to learn in new ways about ourselves as individual sentient beings — and as nations and societies increasingly brought together in an interconnected globe.
Amandeep Singh Gill, India’s Ambassador and Permanent Representative to the Conference on Disarmament, is Chair of the Group of Governmental Experts of the Convention on Certain Conventional Weapons (CCW) on emerging technologies related to lethal autonomous weapon systems
Leave your comments and read more on The Hindu News app