Post in evidenza

AI Revolution: Robot Operation System

Sophia, the first humanoid robot to receive citizenship of a country.   She was featured on the cover of Ella Brazil magazine and has a...

venerdì 2 dicembre 2016

Nightmare Machine

A new artificial intelligence program out of MIT seems like the start to a Black Mirror episode
The university’s Media Lab recently created a new AI called “Nightmare Machine,” which uses deep learning to terrify humans. The algorithm has learned which types of aesthetics disturb and horrify us and has used this to transform regular images into scenes from our nightmares.
“Since centuries, across geographies, religions, and cultures people try to innovate ways of scaring each other. Creating a visceral emotion such as fear remains one of the cornerstones of human creativity,” reads the AI’s website. “This challenge is especially important in a time where we wonder what the limits of artificial intelligence are: Can machines learn to scare us?
So far, the lab has released computer generated scary images in two collections: Haunted Faces and Haunted Places. Some editing styles the team has used include: Inferno, Ghost Town, Toxic City, Slaughter House, Tentacle Monster and Alien Invasion. The AI has transformed places like the Eiffel Tower, New York City and the White House, as well as various unidentifiable faces. To help the algorithm learn, you can go here to vote on which faces you find scary.
Flip through the slides above to see some of the scariest images “Nightmare Machine” has created.

11 Terrifying Images Created by MIT’s New AI, Which Was Built to Horrify Us  • 

Oh Good, MIT Made an AI Nightmare Machine  Rhett Jones 10/22/16 

Dreamception starts with either an image that's been saved on your iOS device or it takes a photo with the camera. You then select one of the thirty available effects and the image is uploaded to the publisher’s server where it’s run through the Deep Dream process and returned. The delay going from mobile device to server and back again is remarkably short so repeating the “dreaming” process and layering multiple effects is fast and easy to do. A nice touch is the vertically split review screen which allows you to compare the original with the "dreamt" image.

Dreamception: Deep Dreaming for iOS Aug 28, 2015

Google’s “DeepDream” algorithm, which uses an Artificial Neural Network connected to itself to form a feedback loop that generates some pretty wild results. Since then, the trippy results of this algorithm has made its way into the bloodstream of the internet and gone fully viral. Happy to encourage some unbridled nerdiness, the Googlers turned around and completely open-sourced the code. Once the nerds got their teeth into it, #DeepDream (which has now been hashtagged) showed off what’s going on in the internet’s id. Unsurprisingly, there are a lot of weird-looking puppies.

And scenes from Mad Max that resemble Hieronymus Bosch paintings

There are way too many fantastic images to showcase here, but I can’t help but include the monstrosity below.  If there’s ever an effective “Hey kids, maybe doing acid is a bad idea” campaign, it’s Fear and Loathing in Las Vegas, as processed by DeepDream.  Happy Nightmares!

#DeepDream’s Greatest Hits

Ti piace?

Nessun commento:

Related Posts Plugin for WordPress, Blogger...

Ratings by outbrain