Is Google's Deep Dream art?
A little while ago, Google set an Artificial Neural Network wild to “dream” on the internet. We asked art critics, artists and artistic technology experts if this was art.
A little while ago, Google set an Artificial Neural Network wild to “dream” on the internet, digging through visual data, “enhancing” parts of images and building on features it “recognizes” within by using its own datasets. (Technical explanation here.) The result was a hazy, swirling glaze of colorful noise and objects mercilessly shapeshifted over and over. Trippy! (Technical explanation here.) It “saw” a lot of dogs. (There are a lot of dogs on the internet.)
But is this art?
Not only is the neural network “dreaming”, it's possibly making art. Or is it? Well, at least now that the “inceptionism” algorithm is freely available online, people have certainly been using it creatively. (Fear and Loathing, Quake, donuts, the universe, sheep...) But as a project, is Google's #deepdream robot an art piece in its own right? Does the very fact that it produces images that are pleasing to humans make the algorithm (or, at least its corporate creator) a sort of an artist? Can you program artistic agency into a bot or an “AI,” or is that idea a logical fallacy? We digress.
“Inceptionism” algorithm as art, much like those #deepdream porn-bug-dicks (NSFW), is a twisted, confusing concept. For our inaugural Is this art? inquiry, we asked art critics, artists and artistic technology experts to share their thoughts.
Nope. It's a tool, not the product, so calling it art would be a little like an artist raising their hand and declaring their paintbrush art because they were so happy with the way they used it lay paint on a canvas. Even the software engineers make this distinction on the Google blog:
Two weeks ago we blogged about a visualization tool designed to help us understand how neural networks work and what each layer has learned. In addition to gaining some insight on how these networks carry out classification tasks, we found that this process also generated some beautiful art.
It's worth noting, though, that in the US corporations are considered people, so Google could deem their software art. If we take as a given that one person's the intent to make art, thus creates a work of art (regardless of its merit) it's certainly a possible future for Google. It sounds a bit like a sci-novel though, so I'm guessing we won't see the corporation do that any time soon.
↑ #deepdream application by Kyle McDonald
I would say there is a good case to say that it is, dependent (as is usually the case with art) on how you 'frame' the idea or interpret the semiotics of process and results.
'Inceptionism' (as it was originally called) came about through discovery through research - the simplest description is: A neural network was set up to interpet images for computing machines to recognize objects in them, and gather this learnt knowledge, a dataset. This idea was then put in reverse: knowledge of what objects looked like could be used to locate and generate their appearance if it could see them in an image.
It is easy to dismiss this as a modern form of psychadelia, but what this has made accessible to the public the idea of an artificial psychology of an artificial intelligence, and how it can connect to us: humans and computers are pattern-recognition entities processing immediate information. Ideas such as pareidolia, habituation and the 'Tetris Effect' come to mind, how our repetitive mental actions leave their impressions on us. If art has the purpose to illustrate the human condition, #deepdream has become a high profile entry point in this field.
Releasing the source code has brought some fun experimentation for creative coders (not seen since the release of the Microsoft Kinect, which itself has been a modern staple for tech art). It could be considered a PR move from Google itself, based on the success of the original Inceptionism post, and we can quickly get tired of the formulations based on the dataset (the constant stream of images featuring 'puppyslugs' will eventually bore everyone), but the subject of Neural Network Art is developing and what is interesting is where it can go next. Examples include 'LSTM' by Sebastian Schmieg which is an app which generates new texts from the works of Ray Kurzweil. Creative coder Samim has produced several generative projects using neural networks to create machine-generated Obama political speeches and TED Talks, or how computer vision interprets pornography. Matthew Plummer Fernandez has created a bot which utilizes deep learning to interpret pieces of artworks, and posts its impressions onto a Tumblr blog.
↑ #DeepDream Experiments, Anthony Antonellis
Google Deep Dream is a medium. On its own it's not art, but the images it’s being used to create can be art. Fugly art.
It reminds me of the generative fractal computer art from in the 80s that filled up the columns of my grade school textbooks. Some of the results look like trippy scenes that could be used in a Pixar version of Fantasia. Conceptually it is very interesting; aesthetically it looks like visual Morgellons disease. I'm sure there could be some compelling results; it seems similar to the Photoshop Content-Aware Fill trend. I’m always happy to see a medium that so much of the public enjoys experimenting and playing with, I just wish those results would get printed out and put on their fridges rather than vomited onto my feed like the mid-2015 version of duck face.
Google Deep Dream is our punishment for not liking Google Glass.
Short answer: Of course it's art! There's no limit to what you can classify as "art." The question is only ever whether it's good art. And people seem to be very amused by it.
Longer answer: It's not art. The way it is getting used is essentially like a psychedelic Instagram filter, and the results are actually a bit repetitious, don't you think? There were already very, very striking images produced by algorithmic means, so I'm not sure what the hoopla is.
Even so, I don't doubt that you can invent AI that can figure out how to make something that has a lot if not all the characteristics of what we call "art," even the really brainy stuff. There will be a Turing test where you won't be able to tell what is made by a human and what is made by a computer intelligence, no doubt about it.
But really what "art" as a category means to us is an invention of the Romantic period in Europe. And what it tends to mean very specifically is "proof of human creative genius," which took on extra cult-like status in response to industrialization, as people tried to find ways to feel like they were holding onto their humanity in a fast-technologizing world. So, as humans have invented new tools of image-making and form-making—photography being the classic example—what tends to happen is that what we call "art" mutates to find a new way to convey "human creative genius."
So, it's like, "OK, the photographers do portraits; painting is about exploring color and form and expression now," which is what happened about when photography became mainstream at the end of the 1800s. And then, what also happens is that, some other artists figure out how to use the tool to convey the new standard of what "art" is, and you get something like "art photography." And that tends to be how it goes.
Maybe the smarter and more creative our computers get, the harder it gets for artists to find new strategies to symbolize "human creativity." Maybe the idea of celebrating exceptional "human creativity", in fact, is dated. But I'm pretty sure that's what the cult of "art" and the cult of the "artist" means, even today, so in that sense, Deep Dream just represents a modest new displacement, or challenge that artists holding onto that tradition have to tangle with, that's all.
Cover image: New Aesthetic