I guess I should warn you that some of the images and gifs in this post are a bit ... disconcerting. Like this one:
This image, which was recently on the front page of Reddit (to the internet's general horror), is the result of an experiment with Deep Dream, Google's neural network. Despite the title and what some on the internet would have you believe, no, this is not what a computer's 'dreams' look like. In fact, no computers were dreaming in the creation of this gif.
What did happen was a lot more simple and more interesting. As explained in this Reddit "Explain Like I'm Five" thread, Google has highly-developed image recognition software. If you give it an image and tell it to look for things, it will compare the image to many others and figure out what's there. But, you have other options: What if you force the software to look for dogs in a picture that has no dogs? Well, it will start identifying even the vaguest dog-like shape as a dog.
And if you start letting the program change the shapes in the image to fit what it's looking for, and then you start feeding it back the results, well, then you get a plate full of spaghetti-dog, as above.
And it doesn't have to be dogs, it can try to find pretty much anything. Check out this video with a series of Edvard Munch's The Scream processed to try to find lots of different things:
And now Google has opened up the code to do this to the public, allowing anyone to take their own images and turn them into Dali-esque nightmare-scapes to share with the world. What fun!
Wired has a running collection of these kinds of pictures that are worth checking out if you're morbidly curious. Some of the images put through this process can actually be aesthetically pleasing, under the right circumstances:
But most of them are just surreal in a bad-acid-trip sort of way, like when someone decided to put a few minutes of the movie Fear and Loathing in Las Vegas through the process.
If you watch it, you can see where the neural network thinks there's a shape and tries to force it in, like where a man's arm, bent at the elbow, becomes the mouth of an animal of some sort. Also, the floor seems to jump from bird to gecko to fish.
I share this news partly because I like sharing weird things with people, but also because I've written before about automation and what complex algorithms can do when dealing with ambiguity. We're going to be seeing a lot more of this stuff as our programs and software grow more automated and computers start tackling the unknowns of our world. I hope your ready for it. Also, here's a 'puppy slug' nebula to send you on your way:
Or maybe you'd like a gif of someone walking in sand ... and maybe eyeballs I don't know:
Oh, and if you think it's upsetting having to see these images, how do you think I feel? I'm the one who actually had to write this post.