Editorial

DARPA Dweebs Spent 6 Days Training A Robot To Detect Humans, Then 8 Goofy Marines Trounced Their AI

(Photo by Lintao Zhang/Getty Images)

Sarah Wilder Social Issues Reporter
Font Size:

A group of Marines managed to trounce a robot after scientists from the Defense Advanced Research Projects Agency (DARPA) spent almost a week training it to detect humans.

The story was told in a soon-to-be-released book titled “Four Battlegrounds: Power in the Age of Artificial Intelligence,” by defense expert Paul Scharre. The volume explores “the struggle to control artificial intelligence.” (RELATED: Police Arrest Suspect After Video Showed Attempted Kidnapping Of Barista At Drive-Thru Window)

The DARPA team had the Marines walk around as they trained the robot to detect human beings. After six days, the Marines were told to try to outwit the AI system.

“They parked the robot in the middle of a traffic circle and the Marines had to approach it undetected starting from a long distance away,” Scharre writes in his book.

The Marine who managed to get up to the robot and touch it without being detected would win the game. But all eight Marines managed to do so with “clever tricks.”

“Two somersaulted for 300 meters; never got detected. Two hid under a cardboard box,” one member of the DARPA team said. “One guy, my favorite, he field stripped a fir tree and walked like a fir tree. You can see his smile, and that’s about all you see.”

Artificial Intelligence is the new scientific advancement that has technology circles buzzing. Researchers at Meta taught artificial intelligence how to play a strategic game. When the robot played against humans, it scored more than double the average human player’s score.

The World Economic Forum (WEF) proposed using the technological capability to identify “misinformation” in August.

“Supplementing this smarter automated detection with human expertise to review edge cases and identify false positives and negatives and then feeding those findings back into training sets will allow us to create AI with human intelligence baked in,” ActiveFence Trusty & Safety Vice President Inbal Goldberger, the author of the article on WEF’s website, said.