Tired of that smug look of satisfaction on your kid’s face when they’re able to find Waldo on a page faster than you can? A creative agency called redpepper built a robot that levels the Where’s Waldo playing field using a camera and machine learning AI to spot the striped traveler in as little as four-and-a-half seconds. Looks like you’re out of a job, little Billy.
Once the robot arm is in place, it snaps a high-res photo of a two-page Where’s Waldo spread, and analyzes it for recognizable cartoon human faces using the OpenCV computer vision programming tools. To actually spot Waldo, the robot’s designers trained Google’s Cloud AutoML machine learning tool on a relatively small collection of 107 Waldo head and body images.
Thousands of sample images are usually needed to properly train an image recognition AI, but because the robot is only dealing with simple cartoons of people, it can pick out Waldo faster than most kids—and easily dominate adults. When Waldo is spotted, the robot arm lowers a creepy silicone hand to point out the no longer hidden character. It might not be the most functional or practical robot ever built, but it’s another small step toward improving the capabilities of machine learning AIs and another example of how the days of humanity’s superiority are numbered.