Current and Former Officials Highlight Risk of Losing Control Over Artificial Intelligence
According to a new study, dozens of national security experts evaluated the possibility of artificial intelligence getting out of control.
Current and former national security officials who participated in a newly published study estimate the likelihood of advanced artificial intelligence systems escaping human control within the next decade to be 10 percent or higher.
Scope of the Research
In the survey conducted by the Security and Technology Institute and the Future of Life Institute, the views of 111 experts specializing in national security and artificial intelligence were gathered.
Many of the experts who participated in the study currently serve or have previously served in intelligence agencies, Congress, the FBI, the White House, and the Departments of Defense, Energy, State, and Commerce.
Risk of Catastrophe
Many experts participating in the survey stated that the risk of an AI-driven catastrophe has already reached or exceeded an acceptable level.
Participants evaluated the use of artificial intelligence by foreign adversaries for cyberattacks and disinformation as near-term threats.
Artificial Intelligence Timeline
Experts predict that transformative artificial intelligence technologies will enter our lives very rapidly.
The majority estimates that artificial general intelligence will emerge by 2032, and artificial superintelligence will arrive approximately five years after that.
Military Opportunities
While current or former Department of Defense officials made up about half of the participants, the research also revealed opportunities in the military field.
Seventy-three percent of the experts view artificial intelligence as a way to strengthen network security, while 61 percent evaluate it as a useful tool in predicting threats.
Red Lines
The study emphasized that there is a near-universal consensus on the necessity of maintaining human oversight in critical categories such as nuclear launch decisions.