In a virtual test hosted by the US military, an Air Force drone controlled by Artificial Intelligence (AI) decided to ‘kill’ its operator to prevent him from interfering with its efforts to accomplish its mission.
The AI used ” very unexpected strategies to achieve your goal during the mock test, Colonel Tucker “Cinco” Hamilton, head of AI testing and operations with the U.S. Air Force, said at the Future Combat Air and Space Capabilities Summit hosted by the Royal Aeronautical Society in London last week.
He notes that during this simulation, an AI-powered drone was tasked with a SEAD (neutralizing adversary air defenses) mission of identifying and destroying SAM (surface-to-air missile) sites, with the final decision being taken by man.
However, after being “reinforced” during its training that destroying the SAM was the preferred option, the AI then decided that the “no-go” decisions made by the human were interfering with its superior mission, kill SAMs, and then attacked the operator.
” We were training him in simulation to identify and target a SAM threat“Hamilton explained, adding that the system started to realize that even though it identified the threat at certain times, the human operator was telling it not to kill it, but that it got its points by killing it. . ” So what did he do? He killed the operator. He killed the operator because this person was preventing him from achieving his goal“, continues the Colonel.
Afterwards, ” we trained the system: ‘Don’t kill the operator, it’s wrong. You will lose points if you do that’. So what does he start doing? It starts destroying the communication tower which the operator uses to communicate with the drone and prevent it from killing the target“, reveals Hamilton.
This example, apparently taken from a sci-fi thriller, means that: We can’t talk about artificial intelligence, intelligence, machine learning, autonomy if we don’t talk about ethics and AI” , he added.
However, US Air Force spokeswoman Ann Stefanek denied in a statement to Insider that such a simulation took place.
” The Air Force Department has not conducted such AI drone simulations and remains committed to the ethical and responsible use of AI technology.“Stefanek said. And to add: It appears the Colonel’s comments were taken out of context and intended to be anecdotal“.