KHARKIV REGION, Ukraine, Nov 29 (Reuters) – Ukrainian drone pilot Mex recalled attacking a recent high-value target – and how the 20-kilometre strike on what appeared to be a Russian tank would have been impossible without the help of an AI-assisted targeting system. “Without the additional guidance, we simply could not hit it,” said the […]
World
Ukrainian drone pilots look to AI for battlefield edge
Audio By Carbonatix
KHARKIV REGION, Ukraine, Nov 29 (Reuters) – Ukrainian drone pilot Mex recalled attacking a recent high-value target – and how the 20-kilometre strike on what appeared to be a Russian tank would have been impossible without the help of an AI-assisted targeting system.
“Without the additional guidance, we simply could not hit it,” said the 31-year-old from Ukraine’s 58th Separate Rifle Brigade, using his military call sign. “Absolutely no way.”
Drone technology has evolved rapidly over the nearly four years since Russia invaded Ukraine.
As both countries produce several million drones a year, the use of signal-jamming devices to cut the connection between drones and their pilots has become ubiquitous on the front lines.
DENSE FOG OF INTERFERENCE
Drones like the one used by Mex to hit the tank are increasingly being used to break through the dense fog of interference which has made countless Ukrainian and Russian drones fall out of the sky before they reach their targets.
These drones lock onto the image of a target seen through their onboard camera, and can continue flying towards it autonomously even after the pilot loses contact with the drone.
The technology, which has also been deployed by Russia, has raised ethical concerns, but internationally binding standards on the use of AI-augmented weapons remain virtually non-existent.
Ukraine, which has rolled out dozens of such systems, with some being installed in thousands of drones, has said any decision to strike must be made by a human and not by AI.
Industry figures say the software often still needs improvement and its performance depends on battlefield conditions.
“This guidance system, in addition to tracking the (target), has its own memory bank, which it has been trained on, with motorcycles, cars, and so on,” said Mex.
“Therefore, if, for example, I set a pinpoint 2-3 kilometres away from the car during approach, it will already adjust to it,” he said.
(Reporting by Volodymyr Pavlov; Writing by Max Hunder; Editing by Conor Humphries)