Google Deep Dreams of Austin, TX

Posted by Mary Ann Castro on Friday, July 24th, 2015 at 10:27am.

Ever wonder how Google deep dreams of Austin? Well, we put several of Austin's sights and landscapes through Google's deep dream AI emulator and found out! Here are the results:

Deep Dream 1: Austin TX Skyline

Deep Dream 2: Downtown Austin TX

Deep Dream 3: Texas State Capitol Building, Austin TX

Deep Dream 4: Terry's Texas Ranger at Austin's Texas State Capitol Building

Deep Dream 5: Austin's Texas State Capitol Building

Deep Dream 6: Austin's Evening Skyline

What is Google Deep Dream?

Google Deep Dream is an open-source software that takes an image and then modifies it based on artificial intelligence. It recognizes similar features between images, and then modifies the images to fill in features that it thinks are missing.

For example, Deep Dream 2 has a field in it. Based on Google's image database, images of fields should be filled with people, tractors and cows. The AI dream decides that this field needs to have those things to, so it fills them in.

And while that makes perfect sense, it doesn't really tell us why the Texas Ranger's horse needed to have multiple heads or why vehicles belong under the South Congress Street bridge. In some ways, Artificial Intelligence really does have a mind (and dreams) of its own!

1 Response to "Google Deep Dreams of Austin, TX"

Calimero wrote:
Can we start with the definition of AI first? Association for the Advancement of Artificial Intelligence says that AI is "The scientific understanding of the mechanisms underlying thought and intelligent behavior and their embodiment in machines."

So what you see is not AI - not scientific and not intelligence behavior. Try this: tomorrow you take a picture of the skyline form a different angle and take it through Google Deep dream - you may see a sensibly different set of changes - what does it tell you?

Sweet dreams!
Calimero

is this how you imagine Austin's skyline? Posted on Tuesday, December 29th, 2015 at 8:32pm.

Want to join an award-winning team?

Work with us