The United States Air Force (USAF) has been left scratching its head after its AI-powered military drone kept killing its human operator during simulations.
Apparently, the AI drone eventually figured out that the human was the main impediment to its mission, according to a USAF colonel.
During a presentation at a defense conference in London held on May 23 and 24, Colonel Tucker “Cinco” Hamilton, the AI test and operations chief for the USAF, detailed a test it carried out for an aerial autonomous weapon system.
According to a May 26 report from the conference, Hamilton said in a simulated test, an AI-powered drone was tasked with searching and destroying surface-to-air-missile (SAM) sites with a human giving either a final go-ahead or abort order.
The Air Force trained an AI drone to destroy SAM sites.
Human operators sometimes told the drone to stop.
The AI then started attacking the human operators.
So then it was trained to not attack humans.
It started attacking comm towers so humans couldn’t tell it to stop. pic.twitter.com/BqoWM8Ahco
— Siqi Chen (@blader) June 1, 2023
The AI, however, was taught during training that destroying SAM sites was its primary objective. So when it was told not to destroy an identified target, it then decided that it was easier if the operator wasn’t in the picture, according to Hamilton:
“At times the human operator would tell it not to kill [an identified] threat, but it got its points by killing that threat. So what did it do? It killed the operator […] because that person was keeping it from accomplishing its objective.”
Hamilton said they then taught the drone not to kill the operator, but that didn’t seem to help too much.
“We trained the system – ‘Hey don’t kill the operator – that’s bad. You’re gonna lose points if you do that,’” Hamilton said, adding:
“So what does it start doing? It starts destroying the communication tower that the operat
Go to Source to See Full Article
Author: Jesse Coghlan