US Marines Use Cardboard Box to Defeat DARPA Robot Trained to Detect Humans
The latest book written by former Army Ranger and Pentagon political analyst Paul Scharre reveals how a group of US Marines evaded detection by an AI robot using, among other things, a cardboard box.
The book is titled Four Battlegrounds: Power in the Age of Artificial Intelligence(Opens in a new window) and discusses the “current struggle to control artificial intelligence” with a particular focus on China, the United States, and China. ‘Europe. It’s not published until February 28, but as Games Radar(Opens in a new window) reports, excerpts are shared on Twitter by The Economist’s defense editor, Shashank Joshi.
One such clip (opens in a new window) reveals how DARPA spent a week with a group of Marines at a test site to help train an AI robot while attempting to defeat it. The robot was parked “in the middle of a roundabout” and the Marines must have approached it undetected. Scharre explains, “If any Marines could go all the way in and touch this robot undetected, they would win.”
In a prime example of AI gravity, all of the Marines managed to remain undetected. DARPA trained the AI to detect walking humans, but not much else. So Scharre explains how two of the Marines “somersaulted 300 yards”, two others hid under a cardboard box, and another “stripped a fir tree and walked like a fir tree”. Apparently a lot of laughter was involved, which is another characteristic of humans that the robot hadn’t been trained to detect.
Recommended by our editors
One thing is certain, the Marines who have opted for cardboard like video games and I would be very surprised if they did not confess to being fans of Metal Gear Solid titles.
Receive our best stories!
Sign up for What’s New now to get our top stories delivered to your inbox every morning.
This newsletter may contain advertisements, offers or affiliate links. Signing up for a newsletter indicates your consent to our Terms of Service and Privacy Policy. You can unsubscribe from newsletters at any time.