Robot Art - The Work of Pindar and his Painting Robots, PHD-1 and PHD-2.


    

Though human creativity and artificial intelligence are related in a broad sense, there is currently little crossover between the two. My art deals with exploring each and trying to find the connections. While my subject matter and style is diverse, the unique process behind all my work is similar.

My primary artistic tools are artificially intelligent painting robots named PHD-1 and PHD-2. These robots not only attempts to mimic human painting techniques by using a brush to apply paint to canvas, but they also attempt to interpret and compose artwork of their own. To achieve this they use a number of artificial intelligence algorithms (neural-nets) including but not limited to k-means clustering, back propagation, and max-nets.

Our artwork together can be seen as a collaborative effort. In this collaboration, I supply the inspiration for the paintings in the form of subject matter. The robots then take my seed image, and using the algorithms that I have written for them, interpret the image and produce an original composition. With the original composition in memory, my robots then direct me to mix a paint palette for use in the painting. I set the paint out, put a brush in their robotic arm, and tell them to begin painting. Anywhere between 1-48 hours later a painting is produced. If it looks complete, I sign it. If it does not, I go back into it by hand, sometimes even painting over the entire canvas. Most of my work is a back and forth between myself and my robots. (See the Process Section for more details.)



Video of the painting robot in action.


My art is not just the finished paintings produced by the robots, but the process described above. The finished paintings are simply an artifact of all the steps that went into creating them, solid proof of my collaboration with my robots. In the current arrangement, I provide most of the creative input while PHD-1 and PHD-2 provide most of the technical skill.

I am constantly attempting to increase PHD-1 and PHD-2’s input into the creative process. The most recent improvement was the addition of a real time camera to PHD-1. With this camera, the robot has two new capabilities. The first is that I no longer need to provide input. The robot can be set up to paint whatever is captured by the camera. The second is that it can record its own progress and make adjustments as needed. In essence, it is now capable of the same self evaluation that all artist perform when creating pieces of art. The question, of course, is how well it can perform the self evaluation.



Live image from camera mounted on PHD-1's robotic head (when in progress).


In addition to collaborating with the robots, I like to bring in as many additional collaborators as possible. For example, many of my paintings are collaboration with my two pre-school children. In these paintings, they are the ones that choose the subject matter and paint the background. The robot then paints the selected subject on their backgrounds. Other paintings are a collaborative work between me, a patron, and the robot. And still others are collaboration between my students, other artists, and the robot. For my artwork, I believe the more creative input the better. I love not knowing where a particular work is going. The best way I have found to ensure this is to include as many collaborators as possible, and then in the end throw the art work to the mercy of multiple artificial intelligence algorithms.


Pindar Van Arman