Where are the robots drawing the line? When the blind dogs become murderous? Autonomous Peeping Toms watching your kids? Or film stars that make us cry over the trailers of Transformers?
Perhaps the final straw is when our childhoods start to be ruined, as the Waldo robot appears determined to do.
The Google machine learning – driven robot, built by the creative ad agency Redpepper, uses facial recognition to dominate the classic, analog hidden image game. And it’s kind of rude, honestly.
Waldo finds the sneaky spectacled cartoon man in a whopping 4.5 seconds in a video that seems to mock the pathetic ineptitude of all humanity. After scanning each face on the page, his silicon hand points to Waldo with an implied attitude of “You morons.”
There’s a Waldo robot consisting of a Raspberry Pi computer chip powered robot arm. But the key element of its puzzle – solving mastery is the camera that allows the facial recognition capabilities of the robot.
The process of how Waldo works is really straightforward.
The camera is taking a snapshot of the chaotic mess, a Where’s Waldo? Scenes. Scenes. Then each face is captured, analyzed by the machine learning algorithm, and it was trained to recognize compared to the Waldo photos. Last, the A.I. Choose the face that best suits what you’re looking for, Waldo in this case.
It does all this in mere seconds — apparently with more than 95% confidence. The silicon hand fitted to the end of the robot arm promptly returns to the resting position when the job is done, as if it would dare us challenge it again.
The video shows how easy this fun – ruining robot was to create. Redpepper technologist Matt Reed told The Verge they just had to train Google images with “62 distinct Waldo heads and 45 Waldo heads plus body” to make it that accurate.
Sure, on Where’s Waldo, using machine – precise facial recognition? Puzzle feels like a knife fight with a homing missile. But if it keeps actual knives out of the robots ‘ hands, let Waldo have fun.