An excerpt from an upcoming book has nabbed the internet’s attention by sharing the tale of how a few United States Marines managed to defeat an Artificial Intelligence (AI) system belonging to the Pentagon’s Defense Advanced Research Projects Agency (DARPA). And it wasn’t even a contest.
Author Paul Scharre, a former Pentagon policy analyst and Army veteran, features the story in his upcoming book “Four Battlegrounds: Power in the Age of Artificial Intelligence.” The Economist’s defense editor Shashank Joshi apparently got his hands on an early copy and shared a screenshot of an excerpt from the book on Twitter, where it’s attained over 4 million views.
These aren’t the marines you’re looking for,” Joshi said in his tweet.
According to the book, a DARPA team spent a week with a group of Marines testing a robot designed to be able to recognize humans. At first, the Marines assisted in developing the robot’s human recognition algorithm by walking around it, so that it could become familiar. That process took about six days.
The state-of-the-art robots used by the Pentagon had an easily manipulated weakness, according to an upcoming book by a former policy analyst: Though they’re trained to identify human targets, the bots are easily fooled with the most lackluster of disguises.
Eight Marines placed the robot in the center of a traffic circle and found creative ways to approach it, aiming to get close enough to touch the robot without being detected.
Two of the Marines did somersaults for 300 meters. Two more hid under a cardboard box, giggling the entire time. Another took branches from a fir tree and walked along, grinning from ear to ear while pretending to be a tree, according to sources from Scharre’s book.
Not one of the eight was detected.
“The AI had been trained to detect humans walking,” Scharre wrote. “Not humans somersaulting, hiding in a cardboard box, or disguised as a tree. So these simple tricks, which a human would have easily seen through, were sufficient to break the algorithm.”
Scharre says what he likes about the Marine versus DARPA AI test is that it demonstrates a couple of key ideas: AI systems are brittle, they are currently able to be defeated by “simple means,” and humans have the ability to “quickly develop clever exploits on the fly.”
Mark Van der Veen offers some of the most analytical and insightful writings on politics. He regularly opines on the motives and political calculations of politicians and candidates, and whether or not their strategy will work. Van der Veen offers a contrast to many on this list by sticking mainly to a fact-based style of writing that is generally combative with opposing ideologies.