A few weeks ago, Google created algorithmic feedback loops in the artificial neural networks it uses to power Google Photos’ image-searching abilities. The result was a series of hallucinogenic still images that looked like a trippy mash-up between Neuromancer and The Electric Kool-Aid Acid Test. Now, artist Memo Akten has taken Google’s Deep Dream source code and applied it to a video selfie. The result is an infinitely recursive hallucination of Akten’s face, as it might appear to an advanced neural network that has gone out of its mind on acid.
In the description of his video, which Akten calls “Journey through the layers of the mind,” the artist gives a good overview of what exactly is happening here. An advanced neural network is a linked series of chips, each of one playing the role of a neuron in your brain. Trained on a database of millions of photos, some of these neurons will “light up” when they scan an image or video and see something they recognize. Where things get psychotropic is when you take an image of all those lit-up neurons, then feed it back into the network, ad infinitum.
Which is exactly how Akten’s video was made. The artist explains:
[W]hen the network is fed a new unknown image (e.g. me), it tries to make sense of (i.e. recognize) this new image in context of what it already knows, i.e. what it’s already been trained on…The effect is further exaggerated by encouraging the algorithm to generate an image of what it ‘thinks’ it is seeing, and feeding that image back into the input…This is like asking you to draw what you think you see in the clouds, and then asking you to look at your drawing and draw what you think you are seeing in your drawing.
Depending on your point of view, the results are either beautiful or queasily nightmarish. If you fall into the latter camp, all I can say is if you think this is freaky, you really don’t want to know what Google’s Deep Dream Code does to GIFs.
[via Laughing Squid]