Advertisement

Google Creates Bizarre Psychedelic Images With AI Software Loop

Tech giant Google has unveiled a series of psychedelic pictures created by its artificial neural network - and it calls the results ‘inceptionism’.

The images show how Google’s intelligent software learns and what happens when it makes mistakes along the way.

The AI software has been ‘taught’ to recognise features such as people and animals using millions of photos.

When a picture is fed into the network, the software is asked to identify any features it recognises and then modify the pictures using those features.

For example, if the network is fed a picture of clouds and it thinks that one of them looks like a bird, it will use this information to modify the rest of the picture, introducing a series of bird-based images to the picture.

The modified picture is then fed back into the network to go through the same process over and over again on a loop, resulting in some very surreal pictures.

The fewer clear features there are in the original picture, the more the intelligent software will use its ‘imagination’ to work out what it thinks should be there.

In a blog post by Google software engineers Alexander Mordvintsev, Mike Tyka and Christopher Olah, explain the significance behind the pictures:

“Well, we train networks by simply showing them many examples of what we want them to learn, hoping they extract the essence of the matter at hand (e.g., a fork needs a handle and 2-4 tines), and learn to ignore what doesn’t matter (a fork can be any shape, size, color or orientation). But how do you check that the network has correctly learned the right features? It can help to visualize the network’s representation of a fork.”

Artificial neural networks are commonly used in ‘machine learning’ where the software is expected to fill in the blanks itself.

The image recognition software is already in use the company’s new Google Photos service.