The Machine is a project by Colleen McCulla uses Artificial Intelligence(AI) neural networks to explore the boundaries of human and machine expression. This project is a co-creative system that engages in a process of making with technology. I’m exploring the potential of machine learning, which is a subfield of artificial intelligence where a machine writes its own rules to solve a problem, to make art and transform creative practice. This project asks if hybrid human/machine practices are necessary for future survival. It questions how human creativity and expression will evolve with artificial intelligence. The Machine challenges the viewer to consider who and/or what is capable of expression and authorship and how we may consciously participate in shaping our coexistence with advancing technology.
Artificial Intelligence is technology that solves problems by imitating human thought. Machine Learning is a subfield of artificial intelligence where a machine writes its own rules to solve a problem. In the first part of The Machine I’m teaching a computer to learn how to make collages. To teach the computer to learn, I start by showing the computer many examples to train it to learn how to create a specific type of image. In this case I show the computer every step of my collage process over and over again using thousands of my previous artworks as examples. After the model is trained, it is capable of making its own rules to create it’s own solution to the problem and generate art. I engage in dialogue with a series of neural networks in a semi-automated system, by; conceiving of the idea, learning how the technology works, assembling hardware and software required, controlling what training dataset images are used, dictating the flow of the college process. As I train the models, the models learn and trains me on how to communicate with it through the output images(or lack thereof) that it generate.
GENERATIVE ADVERSARIAL NETWORKS
The Machine uses two different Generative Adversarial Networks to generate artwork using a training dataset of over 2,500 of my original collages. A Generative Adversarial Network simultaneously trains two models; a generative(G) model captures “data distribution” and a discriminator(D) estimates the probability that a sample came from training data rather than G. “The training procedure for G is to maximize the probability of D making a mistake”(Goodfellow, et al). The introduction of a framework that include both the G and D model essentially adds a two-player game component. The concept of two simultaneous models within the network working in this manner reminds me of my mental processing that occurs in my own artistic practice. Although I’m not trying to trick myself into thinking that every new a work I create is something that I’ve already made/seen, I am constantly vacillating between judgment and creation.
The first model “pix2pix” uses a Conditional Generative Adversarial Network(cGAN) to learn mapping from an input image to an output image. A trained pix2pix model will generate an output from any arbitrary input. This model is especially flexible because various models can be layered, allowing for generated images to become input images for other trained pix2pix models (Isola et al).
The second model I use is a Deep Convolutional Generative Adversarial Networks(DCGANs). DCGANs are unsupervised learning models that learn a hierarchy of representations from objects parts to scenes both in the generator(G) and the discriminator(D). The limitations of the DCGAN lays in the image size of 128 X 128 pixels. Additionally, this is an open source model that is available for use on github (Radford et al).
Multi-layer Recurrent Neural Network
Machinifesto text generated using Multi-layer Recurrent Neural Network trained on 750 plain text pages of every piece of writing I completed during graduate school.
Based in Chicago, IL, McCulla’s recent work explores the boundaries between machine learning and human creativity. McCulla is a TEDx speaker who has been featured at SCOPE ART Miami Beach and nationwide on PBS. Colleen also created and hosts a video series, Paper Cuts, which explores the medium of collage on YouTube. Colleen has created one collage everyday since 2012 which can be viewed on her Instagram.
For inquiries email email@example.com