We’re knee-deep in military hardware on the features desk this week, as we prepare our annual defence special issue, and it’s made our thoughts turn to the future of warfare. Of course, we’re helped in our imaginings by innumerable science fiction films of varying quality, and we can only hope that scenarios such as ‘The Terminator’ are hopelessly wide of the mark.
The start of this week saw Professor Noel Sharkey, machine learning specialist from Sheffield University and an old friend of The Engineer, warn of the dangers inherent in the convergence of artificial intelligence and military robotics. Put armaments on a machine, and equip it with autonomous decision-making, and sooner or later it will make a dreadful mistake, he said.
Prof Sharkey believes that AI technology is 50 years off being able to distinguish between friend and foe reliably, and no computer brain can make the decisions necessary for a proportional response. Of course, we already have armed robots in the battlefield, or at least above it — both US and UK armed forces use armed UAVs — but as the rules of engagement can’t be programmed in to their controlled systems, that aspect of their function is handled via remote control.
But is that likely to remain the case? The US Air Force envisages a move to fully-autonomous armed drones by 2047, and we’ve seen many robotic systems under development which, although currently designed for reconnaissance or load-bearing, could conceivably be weaponised. Indeed, it’s difficult to look at a small, stealthy legged robot about the size of a small dog, or an insectoid, segmented climbing machine, capable of scaling a vertical wall, without imagining them carrying explosives.
Prof Sharkey thinks that military strategists’ view of the capabilities of AI is rooted more in the science fiction films I mentioned earlier than in the real world, and he’s concerned that their enthusiasm — and the laudable desire to keep soldiers, airmen and sailors out of danger as much as possible — will lead them towards some of the dystopian futures we see on the screen. Warfare won’t be robot versus robot, it’ll be robot versus human, and that’s not going to have a good outcome.
Stuart Nathan
Special Projects Editor
(Prefers Aliens to Terminator)
Glasgow trial explores AR cues for autonomous road safety
They've ploughed into a few vulnerable road users in the past. Making that less likely will make it spectacularly easy to stop the traffic for...