Research on Neural Network Creates Art

Google’s recently published blog post on its latest findings about neural networks produced some unexpected surreal art. Also called deep learning, artificial neural networks (ANNs) is a group of statistical machine learning models inspired by the biological neural network of the brain. They can be used for functions like image classification and speech recognition, wherein there are usually a large number of inputs, which are also unknown.

Google’s research team concedes that even though ANN applications are already far and wide, there is still uncertainty as to why some models work and some don’t. In its search for answers, Google describes several techniques it uses to get ANNs to automatically generate images based on their understanding of what objects look like. What is unique to Google’s approach (as opposed to research, for example, being done by Facebook) is that it is letting the ANN decide on which visual elements to emphasize. The resulting images are quite intriguing and surreal, so much so that the blog post title “Inceptionism” seems very apt. SEE THE IMAGES HERE.

The visualization techniques presented by Google go a long way into observing and understanding how ANNs process intricate classification tasks and what they may have already learned during training. One interesting point though that the blog post also brought up is this:

It also makes us wonder whether neural networks could become a tool for artists—a new way to remix visual concepts—or perhaps even shed a little light on the roots of the creative process in general.”

You can read more about the techniques mentioned in this article in the original Google post HERE.

Comments are closed.