How? By using simple "cartoon logic" and cardboard boxes during field tests.
In a striking display of low-tech ingenuity, a group of U.S. Marines recently outmaneuvered a sophisticated artificial intelligence detection system developed by DARPA. During a week-long trial, eight Marines were tasked with approaching the sensor without being detected. Rather than using high-tech jamming equipment or complex stealth maneuvers, the soldiers relied on unconventional tactics that seemed to be pulled straight from a Saturday morning cartoon. Two Marines somersaulted for 300 meters, another pair hid beneath a simple cardboard box, and a fifth soldier disguised himself as a fir tree, slowly shuffling toward the target. Incredibly, every single Marine reached the objective without triggering the AI’s sensors.
The failure of the advanced surveillance system highlighted a critical flaw in current machine learning models: the rigidity of training data. The AI had been meticulously trained to recognize the specific movement patterns and gait of humans walking or running, but it lacked the frame of reference to identify individuals rolling through the dirt or mimicking vegetation. Because the Marines' behavior fell outside the narrow parameters of what the algorithm considered "human," the system effectively ignored them. This incident serves as a vital reminder for the defense industry that while AI is an incredibly powerful tool, it remains susceptible to creative human deception and the unpredictable nature of manual tactics.
source: Scharre, P. (2023). Four Battlegrounds: Power in the Age of Artificial Intelligence. W. W. Norton & Company.
Post a Comment