“Deep learning” is a branch of artificial intelligence that allows a computer to learn to identify and categorize data without human supervision, a kind of technology commonly used in image recognition, facial detection, computer vision, and natural language processing software, among many other things. This series, titled Shallow Learning, compares the way people see photographs to the way algorithms see photographs.
Two commonplace yet sophisticated digital tools that recognize or “see” photographs were central to this work. Each of the images in this series is a composite that started as a single image from a past project. These leftover images, which had never been published, exhibited, or posted on the internet, were used as search criteria in Google’s “search by image” feature. This function of the search engine is typically used to track down the provenance of an image found online, but because the images I was searching for did not exist online and had no history, the search engine instead offered a selection of visually similar images-- algorithmic guesses at what these pictures showed. I then selected one of these “best guesses” from Google, placed it next to the original image on a blank canvas in Photoshop, and filled in the area between the two images using the “content aware fill” function.
Though this work is relatively recent (the book was published in 2018), it already feels like a small mark in the historical record of artists navigating the dawn of Artificial Intelligence. Between then and now technology has evolved rapidly. The mistakes and hallucinations the machine embeds while trying to anticipate what we see in a photograph have become more subtle, insidious, and naturalized. Even so, these pictures mark an important stage in that history, when a boundary was being crossed between the world of human meaning making and machine learning.