Here’s what Google’s Artificial Intelligence “sees” when you put a Rorschach test in front of it
Google Images indexes and categorizes trillions of images and it’s not a group of employees sitting there day and night sorting and organizing our jpegs and PNG files.
Google’s algorithm looks at contextual content to help determine the image and how relevant it is but they also have image recognition software at their disposal to ensure that what they’re seeing is in fact, a dog.
Not just that, with millions upon millions of images, they’re able to train their software to recognize traits, details and features of a dog so that the image of your cute puppy you just uploaded to your website is instantly recognized as such.
But not only can Google’s software recognize images, they can also generate images based on those same features it is familiar with.
Now here’s the fascinating part — what if you asked the Google A.I. (or neural network) to use its database of images to not only detect, but to enhance what it recognized, what would you get?
Google engineers decided that instead of asking the software to generate a specific image, they would simply feed it an arbitrary image and then ask it what it saw.
Google’s image recognition software, which can detect, analyze, and even auto-caption images… In a process they’re calling “inceptionism,” Google engineers sought out to see what these artificial networks “dream” of—what, if anything, do they see in a nondescript image of clouds, for instance? What does a fake brain that’s trained to detect images of dogs see when it’s shown a picture of a knight? (from QZ)
Basically, a Rorschach test for Artificial Intelligence. Google’s A.I. came up with a Andy Warhol meets Salvador Dali-like dream world of pig snails and camel birds among other amalgamations:
As you can see, some of the images are beautiful, airy and dream-like while others are more akin to creepy, scary and bloodthirsty night terrors. But that’s just because we associate certain feelings to a dog-waterfall that leads into a small pool of dog eyeballs, to Google’s neural networks and image recogniztion software, it’s just what it sees.
In Google’s blog post, they’re calling it “Inceptionism“. As much as I am interested in learning more about the nuts and bolts, it’s way above what my brain can handle. I’ll just go back to looking up at the clouds… made up of pigsnails, dogfish and mutated rabbits…