#thespacegoat

So google recently released the source-code for their DeepDream project, which is meant to visualize computer neural networks and show us what it is that computers “see” when they’re categorizing/analyzing images based on learned databases.

These databases are indexes of thousands of images that have been sorted into thousands of categories [you know that captchas that ask you to write down the numbers you see or ask “what 4 images are hamburgers?”? Those all help to feed into the database and categorize images].

The DeepDream project even allows the computer to basically inject what it’s learned from these databases into images and create it’s own images, or ~”dreams”~.

Basically it’s a nightmare machine that shows us how computers see and manipulate pictures, and I tested it on Danny Devito...several times.

Here’s another neat example I found online: