AI-Directed Drones

Can AARDvarks curtail caprice?

In April of 1945, a soldier of the US Army Air Force, uses radio signals to guide the landing of an unmanned aircraft (US National Archives)

The only computer program I ever wrote, coded in BASIC on a Commodore 64, has not seen the inside of a CPU for forty years. Likewise, my engagement with artificial intelligence (AI), though much more recent, has been limited to the use of large-language models (such as ChatGPT) to provide second (and third, and fourth) opinions on translations. Finally, my exploration of the issue of self-directed drones consists entirely of the development of a whimsical table-top wargame called We, Robot.

Nonetheless, recent reports of the building, and, indeed, deployment, of human-free aircraft piloted by advanced algorithms convince me that AI-directed drones fall well within the purview of The Tactical Notebook. In particular, I feel compelled to explore an aspect of AI that, if I am not too badly mistaken, promises to pose prickly problems for the wielders of such weapons, something that, faute de mieux, I will call ‘caprice’.

We, Robot

Articles about Drones