Dreaming computers? Google Deep Dream

I recently found out about this weird computer program created by Google called Deep Dream. It is an automatic image editing program that locates patterns in images and superimposes them onto the image to show the user. It's hard to explain what the images look like, so it's good to try it for yourself. Here's a picture that I edited using the service.

What started as a picture of me jumping at my summer job, turned into a trippy representation of the same picture. It's really fun to mess around with different pictures and settings! But what exactly does it do, and how does it work?

It uses a very complex kind of searching called "image recognition". Image recognition is very difficult for computers to do accurately because they are so hard coded to do searches from words. That's why this program is so cool!

As I've talked about in previous blogs, this program uses an artificial neural network (ANN), similar to the one used in IBM Watson, that can learn things as more information is presented to it. Google trained this program using tons of images and repetition to learn what certain objects look like and recreate versions of these objects without further input. For example, after the program is fed many pictures of bikes, it will learn what a bike typically looks like, and can create its own image of a bike on command. This is the kind of technology that was applied to Deep Dream.

Groups of artificial neurons work together to identify different aspects of an image, such as borders and colors, and using this information, try to find other objects, like a bike, that may have the same or similar borders and colors. So when I uploaded the picture of myself jumping, it saw wrinkles in my shirt and colors that created a pattern similar to that of a dog. The neurons overemphasize every aspect of the image, looping through the same process of identifying patterns and accentuating those patterns several times, until a final image is produced. You can even manually upload the same image to the system several times and see what it produces. Here are a few more examples of my own pictures and some that I found on the internet:

The same picture as before, but put through the system about 5 times.

Here's an image before, and after it was put through the system 10 times.

Although this specific program is intended mostly for recreational purposes for the public, the general idea can be applied to technology and programming as a whole. Nobody is telling the computer what to find in these images; the computers have been trained to locate and show what it sees. Some people think that this is a huge step towards artificial intelligence because the computer is creating something on its own that isn't necessarily there. I think this quote sums it up well:

• "It's hard to know exactly what is in control of Deep Dream's output. No one is specifically guiding the software to complete preprogrammed tasks. It's taking some rather vague instructions (find details and accentuate them, over and over again) and completing the jobs without overt human guidance. The resulting images are a representation of that work. Perhaps those representations are machine-created artwork. Maybe it's a manifestation of digital dreams, born of silicon and circuitry. And maybe it's the beginning of a kind of artificial intelligence that will make our computers less reliant on people" (Chandler).

What do you guys think? Also, here's the link to try it out for yourself! There's also a gallery of cool images that people have made through the program. 

https://deepdreamgenerator.com

_________________________________________________________________________________

http://computer.howstuffworks.com/google-deep-dream3.htm

https://en.wikipedia.org/wiki/DeepDream