Prove you're not a robot
Various installation shots:
“Prove you’re not a robot.” This mildly unsettling imperative caught my attention during the spring of 2016. I was surfing the web and, for a reason that I can no longer recall, decided to create an additional Google account. There were the familiar required fields that one may choose to be more or less candid about: name, date of birth, gender (the least flexible option), and location. I was also required to “prove I was not a robot.” In this instance, the proof of my personhood lied in my ability to identify and retype a series of numeric digits that appeared in the middle of a random, poorly compressed and tightly cropped digital photograph that showed a street address from a seemingly anonymous building. The option to refresh the image was also provided and out of curiosity, I chose this option. I refreshed the image, repeatedly, and each time a new fragment containing a street number appeared. These images were uniformly cropped and the series of numbers generally occupied the central third of the horizontal picture plane. In general, the images were formally abstract, though occasionally recognizable features from the world of humans and nature infiltrated their narrow frame: flower pots, tree branches, lawns, windshields, rims, flags, holiday decorations, and other things that robots would have little use for.
The process outlined above, this inverted Turing test, is called “reCAPTCHA.” reCAPTCHA was developed by computer security researchers to prevent bots from exploiting services provided by websites and social media platforms. In a previous iteration, reCAPTCHA used distorted sets of black letters and numbers over a white background as its visual test. At the point at which I encountered the system, it was utilizing photographs of structures and facades extracted from Google Street View, Google’s effort to document every inhabited location on Earth. Over the course of the weeks that followed my initial encounter with this system, I continued to refresh and manually save reCAPTCHA images — thousands of them — to my computer.
Apart from my initial visual curiosity and the slight joy I derived from the surprise of refreshing the reCAPTCHA images, what also intrigued me about these peculiar digital fragments is how they had become operationalized and what this process could reveal about the social, economic, and technological forces that utilized them. On the “surface” the reCAPTCHA images are an element of an automated gate monitoring system, one intended to prevent bots from entering secure access zones by simulating human agents. But they are also linked to a sophisticated machine learning process to train a program to more accurately “recognize” assigned visual markers, in this case, the numeric digits 0-9. It is this latter function that also transforms our computer monitors and keyboards into temporary stations of a modulating and distributed “factory line,” surreptitiously making us digital laborers. None of this is a secret. Google, to their credit, is relatively transparent with regard to their motives: “reCAPTCHA makes positive use of this human effort by channeling the time spent solving [CAPTCHA images] into digitizing text, annotating images, building machine learning datasets. This in turn helps … solve hard AI problems.”
It is this confluence and confusion of categories and relationships – of labor and leisure time; of human beings taking part in the training of systems that appear destined to replace them as workers; of the tacit acceptance that public space may be freely “harvested” for the private economic requirements of a global tech giant; of the vague, yet familiar sense of “placelessness” that imbues these pictures; of the fact that we contribute, willingly, to our own surveillance through the process of creating searchable online profiles of ourselves; and that these ostensibly mundane JPEGs could play a role in “proving” our personhood to a computer – that are encapsulated in the reCAPTCHA images and why I have chosen to use them as an artistic material.
Yet, soon after I began this project, the reCAPTCHA images outlined above were abandoned; researchers concluded that the algorithm used to identify and differentiate numeric digits had achieved functional parity with its human model. Another step had been taken in the process of abstracting human vision in order to supersede it. Google now uses more cunning, furtive methods to differentiate humans from bots, and these reCAPTCHA images, in their obsolescence, have all but vanished from the Internet. Suddenly, the folder that I stored these tiny pictures of the world in began to resemble a digital excavation site.
- E.J. 2017