Artificial Intelligence in Weapon Systems Committee Report Debate
Full Debate: Read Full DebateViscount Waverley
Main Page: Viscount Waverley (Crossbench - Excepted Hereditary)Department Debates - View all Viscount Waverley's debates with the Ministry of Defence
(4 months, 1 week ago)
Lords ChamberMy Lords, “proceed with caution” is for an ideal world, but with warfare on the horizon, it is important to move on from abstract and procedural. With the world headed into a dark place, the geopolitical implications of autonomous weapon systems in modern warfare are immeasurable and will require crucial global diplomacy.
The race for AI supremacy and the increased speed of warfare with first-mover advantage, armed with automated systems, drones and predictive analytics, have implications for the balance of military power between states, even transcending national states, with a far-ranging impact on global peace. Non-state actors that also have access, by one means or another, to this advanced technology will have to be added into the conundrum.
Weapon systems that involve no human oversight present challenges but also opportunities beyond ethical questions. They will test democracy and geopolitics and will change the nature of warfare, in that placing human soldiers in harm’s way will become untenable. Autonomous unmanned underwater, surface and air weaponry can also be set to perform the same tasks of automatically engaging incoming missiles. This becomes machine-speed warfare, with humans no longer the central lethal force in the battlefield.
This is closer than many anticipate. Defensive and offensive AI-controlled fighter jets will become smaller, far faster and more manoeuvrable, and will be able to operate in swarms. Predicting intelligence behaviour with the deployment of kinetic forces against a third party that is not human will accelerate, with the rights norms to proportionate military response in effect no longer applicable.
At state level, therefore, countries should work to establish conventions in which the use of lethal force or nuclear is always subject to human command and control mechanisms, never automated, with emergency communication trip-wire channels or early warning system activates established. As immediate retaliatory responses may no longer be legally or morally justified, the historic conventions of war will require revision, with alternate arbitration systems devised at UN Security Council level.
I have three conjectural questions in conclusion. If an AI system were to physically operate on a human, to what extent should its algorithmic programming be open to public due diligence? Who would be liable in the case of misuse when human oversight is required? How do we counter the spatial distance of a development team at the far end of the world from unethical behaviour, making accountability impractical?
It would be amiss of me to end by not thanking the noble Lord, Lord Lisvane, and his committee.