Email
Password
Remember meForgot password?
    Log in with Twitter

article image'Inceptionism' — Taking a look at the computer's dreams

By Karen Graham     Mar 6, 2016 in Technology
If a computer could dream, what do you think its images would look like? This is the idea behind a project Google began in an effort to understand how artificial intelligence interprets the world.
We all have dreams, some we remember and many more are fleeting images that are forgotten as soon as we awaken. The folks at Google began a process in June last year they dubbed "inceptionism," in an effort to understand how artificial intelligence (AI) interprets the world.
Basically, Google engineers were intent on finding out how AI neural networks carried out classification tasks so engineers could further improve the system. The neural networks had to be trained to identify images, and as they "learned," the network began to gradually adjust its parameters and learned to recognize visual patterns.
A Vincent van Gogh-inspired Google Deep Dream painting created by Google engineers.
A Vincent van Gogh-inspired Google Deep Dream painting created by Google engineers.
Sarah Cascone/Twitter
As an example, let's take Vincent van Gogh's Starry Night. The network identifies a particular image in its data, and operators can feed it additional data to encourage the network to work on that perception. So when the computer thinks part of the sky in Van Gogh's painting looks like a bird, well, the human team runs with it.
A group of 29 paintings made by Google AI were sold at a charity auction in a historic movie theater in the very hip Mission neighborhood of San Francisco on February 26, reports the Wall Street Journal. Believe it or not, but the priciest painting went for a cool $8,000. Not bad for a neural network's first outing in the art world.
Because the project created so much interest with programmers and artists, Google open-sourced its code, called Deep Dream, so anyone could make their own psychedelic paintings. Here are a few examples of how Deep Dreams works:
Untitled
Google
In the above image, you can see a breakdown as to how it works. In the second column, we see Google fed its neural network an image of a tree. Because that neural network is trained to look for buildings in an image, it spat back a squat, green building. Basically, the resulting image depends on what the neural network has learned to look for.
If you think this new inceptionism artwork is only a fad, think again. It's going to become more common. The University of London is now offering a course on Machine Learning and Art, and NYU is offering a similar course of study, says Business Insider.
For those of you who would like to try your hand creating your own Deep Dream piece of artwork, you can go to Google's online Deep Dream Generator. The first picture below is an image of the Golden Gate Bridge in San Francisco.
The Golden Gate Bridge sways  even though we may not be aware of the movement.
The Golden Gate Bridge sways, even though we may not be aware of the movement.
Steven Pavlov
This next image was created on Deep Dream Generator tonight. It is so cool.
I am calling this creation Bridge of Dreams.
I am calling this creation Bridge of Dreams.
Karen Graham
More about inceptionism, computer dreams, Artificial intelligence, Google, neural networks
More news from
Latest News
Top News