Earlier this month, we ran through the workings of Google’s Neural Network “Deep Dream,” a layered algorithm system programmed to function like the neurons in our brains. The goal is to create a system that can recognize a specific subject, say a bird, in any image that graces Google’s scope.
But what’s been mesmerizing the internet is what happens when programmers tell Deep Dream to stop at a certain layer, like one that recognizes animal faces, and repeat the process in a feedback loop. The result is a vom-inducing swirl of color and shapes, lending to the program’s nickname, “The Acid Trip Generator.” It’s weird enough when we’re looking at still photographs of animals, but what happens when you throw moving human faces in the mix? The end of sleeping soundly, that’s what.
This insanity comes way of a GitHub user who melted our brains by running the Hunter S. Thompson flick, Fear and Loathing in Las Vegas, through the network, taking us on a movie trip through the trip-movie (so meta). What you’re seeing here is the generator’s interpretation of each frame of the video as it scans for patterns it can recognize.
Should you want to dive deeper into the hell hole, you can actually do this at home, thanks to these handy instructions. Give your Facebook friends a break by posting the Magic-Eye version of your baby, turn your best friend into Allen Parish getting sucked into Jumanji – the surrealist world of Deep Dream is your proverbial oyster. I shudder to think what Beetlejuice would look like under this treatment.
“It’ll be interesting to see what imagery people are able to generate using the described technique,” say the code’s creators. “If you post images to Google+, Facebook, or Twitter, be sure to tag them with #deepdream so other researchers can check them out too.”
The tweets are already pouring in. Here are some of our favorites creations so far:
— Framestore (@Framestore) July 7, 2015
— Chris Priestman (@CPriestman) July 7, 2015
Just a sec. We need to lie down.