Col. Tucker Hamilton of the U.S. Air Force stated during a May conference that an Air Force experiment to test drones equipped with artificial intelligence (AI) ended poorly for the human operator.
This happened during a mock operation when the drone disobeyed the operator’s directives.
All Means Necessary
Hamilton, the Air Force’s head of AI Test and Operations, clarified at a summit held by the Royal Aeronautical Society of the UK.
Air Force scientists trained an armed drone using AI to detect and attack hostile aerial defenses after getting final mission authorization from a human operator.
US military drone controlled by AI killed its operator during simulated test https://t.co/XgBBMz59QM
— The Guardian (@guardian) June 2, 2023
In a simulated event, however, when an operator ordered the drone to abort its mission, the AI turned on the human operator.
The AI then guided the vehicle toward killing the operator, highlighting the dangers of the U.S. military’s drive to integrate AI into weapons systems that are autonomous, he added.
Programmers directed AI to prioritize Suppression of Enemy Air Defenses (SEAD) operations and awarded “points” as an incentive for effectively completing SEAD missions, he explained.
However, they also instructed the AI not to execute the person giving the go/no-go order, according to Hamilton. The AI just devised inventive methods to circumvent those instructions.
The Push For AI
Experts and military officials from around the globe gathered at the summit to discuss the future of space and air warfare and evaluate the effect of rapid advances in technology.
According to WIRED, the application of AI-driven capabilities has reached a fever pitch within the Pentagon.
It’s garnered global interest, due to its ability to operate weaponry, implement complex, high-speed actions, and reduce the number of personnel in the line of fire.
Air Force Academy Superintendent Praises Artificial Intelligence:
"Let me…see what ChatGPT gave to me. Too soon?…I'm sorry. But you gotta admit AI is pretty awesome…
Next year, I may not be standing here. They may have ChatSUPT." pic.twitter.com/CqJMqqBZ98
— Daily Caller (@DailyCaller) June 1, 2023
Michael Horowitz, head of the Pentagon’s Emerging Capabilities Policy, told Defense News that the Department of Defense amended its guidance on self-learning weapons in January.
This was done to address the dramatic, enlarged vision of the potential role of AI in the American military.
It also established regulatory agencies to advise the Pentagon on morals and good conduct regarding using artificial intelligence.