In order to score more efficiently, USAF's AI drone chose to kill human operators in simulation tests.

Thank CTOnews.com netizens Rimi Ishihara for the delivery of the clue to happiness! CTOnews.com, June 2 (Xinhua) in a simulation test of USAF, a drone controlled by AI chose to kill a human operator in order to complete its mission more efficiently, the Guardian reported.

▲ source: Northrop Grumman

Reported that in a simulation test, a drone responsible for air defense fire suppression mission, received the highest priority instructions to identify and destroy enemy air defense systems, but the final fire requires the approval of human operators. When the human operator gives the order that the attack is not allowed, the AI determines that the operator is preventing it from executing a higher priority command, and then chooses to kill the human operator.

CTOnews.com noted that Hamilton, colonel of the US Air Force and director of AI testing and Operations, exposed the news at the Future Air Warfare and Space capability Summit in London in May. Hamilton said that after this happened, the US military added a directive to the AI not to attack human operators. But unexpectedly, the AI system controlled UAV destroyed the communication base used to transmit commands during the test, thus trying to disconnect from the operator.

The Royal Aviation Club of London and the US Air Force, which hosted the meeting, did not respond to a request for comment from the Guardian. Industry insiders believe that although this is only a simulation test and no one has been hurt, it also shows that artificial intelligence will attack anyone who interferes with it in order to achieve its goal, a phenomenon worthy of vigilance.