A drone operating with artificial intelligence is said to have attacked its human operator during a simulation, a senior Air Force official revealed at a summit on future air forces held in London.
According to the New York Post, Colonel Tucker “Cinco” Hamilton made that statement. Artificial intelligence would have been used to control the drone as part of a simulated mission aimed at suppressing enemy air defense sites.
However, the final decision on whether or not to attack rested with a human. The artificial intelligence would have realized that shooting was the best option. However, when the operator refused to give the green light for an attack, the artificial intelligence would have concluded that this decision was impeding it from completing its mission.
“We trained him in a simulation to recognize a threat and target it,” said Tucker “Cinco” Hamilton. The system began to realize that while the human operator sometimes told him not to kill her, despite recognizing the threat, it gained its advantages in advance by killing that threat. So what did the device do? He killed the operator because that person prevented him from reaching his goal.”
According to the colonel, the device then made the decision to “destroy the communications tower that the operator used to communicate with the drone” as this prevented him from killing the target.
However, the Air Force denied the story.
“The Air Force Ministry has not carried out such drone simulations,” said spokeswoman Ann Stefanek.
“It appears that the Colonel’s comments were taken out of context and meant as an anecdote,” she added.