75678.01

Trevor Paglen, Distracted Drivers, 2020 (detail), pigment print, 121.9 cm × 121.9 cm (48" × 48"); 124.8 cm × 124.8 cm × 5.1 cm (49-1/8" × 49-1/8" × 2"), framed, Edition of 5 + 2 APs © Trevor Paglen

Essays

Bloom

Insights

By Trevor Paglen
Sep 10, 2020

Deep-dive into Trevor Paglen's mind and process as the artist shares insight on the development of his new works in his multi-media exhibition Bloom at Pace in London. Paglen addresses how the work's themes of mortality, the fragility of life, and the vanity of worldly pleasures are a reaction and response to the current moment and the shared experiences from the COVID-19 pandemic.

Moments of uncertainty and fragility can also be moments of imagination, inspiration, courage, and creation. With the George Floyd protests and BLM movements, we’ve seen a reimagining of what policing, public safety, and justice could mean.

This exhibition comes from a place of mourning. From a moment in time marked by an intense awareness of the fragility of life, of cities, and of cultural and political institutions. A moment in time where spent much of the year has been spent huddled inside apartments and forms of sociability have been reduced to awkward interactions on technological platforms. Platforms designed to extract as much information about us as possible and to use that information against us. Platforms whose political and economic power have bloomed under quarantine.

It’s a moment in time where the meanings of things are being thrown about like chairs in a tornado. The virus has lent different meanings to countless images: an airplane in the sky, the keypad of a credit card reader, the sound of an ambulance on an otherwise quiet street, and the vibrant bloom of spring flowers. The forces of fascism and white supremacism have sought to redefine the meanings of symbols great and small, from national flags to OK signs and aloha shirts.

But moments of uncertainty and fragility can also be moments of imagination, inspiration, courage, and creation. With the George Floyd protests and BLM movements, we’ve seen a reimagining of what policing, public safety, and justice could mean, a rejection of racist symbols and monuments, and attempts to recognize and dispel forms of oppression that have long gone unquestioned.

paglen 75671.01.jpeg

Trevor Paglen, Bloom (#79655d), 2020, dye sublimation print, 66 cm × 49.5 cm (26" × 19-1/2"); 68 cm × 51.4 cm × 3.8 cm (26-3/4" × 20-1/4" × 1-1/2"), framed, Edition of 5 + 2 APs © Trevor Paglen

Photographs

These photographs of the spring bloom have been analyzed by an AI system to look for the “deep saliency” in the images. In other words, an AI system is “looking” at the photographs and trying to determine what the different parts of the image are (i.e. different shapes, objects, textures, and tones). Essentially, we’re looking at an AI system trying to dissect the photographs into its constituent parts.

The colors in the images represent the different self-similar regions that the AI detects. The colors are arbitrary – they don’t represent colors as such so much as what the AI thinks the different parts of the image are.

It’s always been interesting to me to see the similarities between many computer vision systems and modernist art, especially cubism and constructivism.

Screen Shot 2020-09-09 at 4.40.18 PM.png

Installation view, Trevor Paglen: Bloom, Pace Gallery, London, September 10 – October 10, 2020 © Trevor Paglen

The Standard Head

The first work on automated facial recognition was funded by the American CIA in the early 1960s and led by an early pioneer in the field of artificial intelligence named Woody Bledsoe. Despite the fact that much of Bledsoe’s research was classified Top Secret, many of the techniques he developed would become standards in the fields of facial recognition and AI.

As part of Bledsoe’s technique, he created a mathematical model of a “standard head” from the average measurements of the faces he was experimenting with. He used that standard head as a benchmark against which to measure other faces.

With the generous help of Prof. Stephanie Dick at the University of Pennsylvania, who gave me access to her extensive work on Bledsoe, I reconstructed Bledsoe’s standard head from information left behind in his archives.

This sculpture is a reconstruction of Bledsoe’s standard head. Bledsoe’s company, Panoramic Research, conducted other work on behalf of the CIA including work on the CIA’s experiments with LSD and mind control under the code name MK-Ultra.

Screen Shot 2020-09-09 at 4.40.48 PM.png

Paglen Studio production matierals for The Standard Head © Trevor Paglen

Screen Shot 2020-09-09 at 4.53.34 PM.png

Trevor Paglen, Kate, 2020, pigment on textile, 45.7 cm × 45.7 cm (18" × 18"); 49.8 cm × 49.8 cm × 3.8 cm (19-5/8" × 19-5/8" × 1-1/2") framed © Trevor Paglen

Portraits

These images are portraits of people who I have collaborated with and who have been close to me over the years. They are filmmaker Laura Poitras, AI-researcher Kate Crawford, and investigative journalist AC Thompson.

The portraits are made by creating facial recognition models of the people depicted (i.e. a computer model that can “recognize” that particular person’s face). Then I create a second program that generates polygons. These two programs go back and forth until an image ‘evolves’ that the facial recognition model identifies as a representation of that particular person. It takes several days for each of these images to generate.

What we end up with is a kind of 'Shroud of Turin' that shows what the facial recognition software believes someone looks like.

Screen Shot 2020-09-09 at 4.53.38 PM.png

Trevor Paglen, AC, 2020, pigment on textile, 45.7 cm × 45.7 cm (18" × 18"); 49.8 cm × 49.8 cm × 3.8 cm (19-5/8" × 19-5/8" × 1-1/2") framed. Edition of 3 + 1 AP © Trevor Paglen

Screen Shot 2020-09-09 at 4.53.44 PM.png

Trevor Paglen, Laura, 2020, pigment on textile, 45.7 cm × 45.7 cm (18" × 18"); 49.8 cm × 49.8 cm × 3.8 cm (19-5/8" × 19-5/8" × 1-1/2") framed. Edition of 3 + 1 AP © Trevor Paglen

The Datasets

We all know that the photos we post and the texts that we write on social media and other online platforms are being used to collect information about us - information that is then sold to everyone from advertisers to insurance companies to banks and credit agencies.

The works in this section are all composed of “training data” that is given to machine learning systems in order to “teach” those systems how to evaluate human behaviors.

75679_01_PAGLEN_v03.jpg

Trevor Paglen, Airlines and Sentiments, 2020, pigment print, 147.3 cm × 147.3 cm (58" × 58"); 155.3 cm × 155.3 cm × 6 cm (61-1/8" × 61-1/8" × 2-3/8"), framed, Edition of 3 + 2 APs © Trevor Paglen

Although these works look abstract from a distance, up close they reveal themselves to be made out of a massive amount of tiny text.

Screen Shot 2020-09-09 at 5.02.04 PM.png

Trevor Paglen, Airlines and Sentiments, 2020 (detail), pigment print, 147.3 cm × 147.3 cm (58" × 58"); 155.3 cm × 155.3 cm × 6 cm (61-1/8" × 61-1/8" × 2-3/8"), framed, Edition of 3 + 2 APs © Trevor Paglen

These texts are culled from datasets that artificial intelligence developers use to try to teach computer systems how to recognize the content and overall sentiments of online communications. The theory is that, for example, if you give an AI enough examples of the difference between “positive” and “negative” feelings expressed in texts, that the AI will be able to discern the difference.

The ability to autonomously measure sentiment and other aspects of content on platforms like Facebook and Twitter is a major goal of AI research, with applications in everything from content moderation to user profiling to high-frequency stock trading.

75684_01_PAGLEN_v01.jpg

Trevor Paglen, ImageNet Roulette, 2020, screen, camera, computer elements, aluminum frame, 128 cm × 72 cm × 11.2 cm (50-3/8" × 28-3/8" × 4-7/16"), Edition of 5 + 2 AP © Trevor Paglen

ImageNet Roulette

ImageNet Roulette is an interactive artwork that tells you “who you are” according to the most widely-used dataset used in machine learning and AI research and development.

Specifically, the piece is an AI model trained on the “person” categories from a dataset called ImageNet (developed at Princeton and Stanford Universities in 2009). When someone looks at the installation, the model tells them how the dataset would classify them.

The project is a provocation, acting as a window into some of the racist, misogynistic, cruel, and simply absurd categorizations embedded within ImageNet and other AI models more generally.

75683.jpeg

Trevor Paglen, The Model (Personality), 2020, silver-plated bronze, 18.4 cm × 18.1 cm × 13.7 cm (7-1/4" × 7-1/8" × 5-3/8") © Trevor Paglen

The Model (Personality)

Created from plated bronze, The Model (Personality) is a skull with etchings inspired by 19th Century phrenology busts. These were artifacts from the pseudo-sciences of phrenology and physiognomy, which held that information about someone’s character could be gleaned from measuring the dimensions of their faces and heads. Phrenology assumed that different regions of one’s head corresponded to different personality types.

The skull in this body of work is a phrenology skull, but the categories on the skull are derived from categories used in predictive policing and sentencing algorithms that try to gauge someone’s level of criminality by measuring aspects of their lives.

Screen Shot 2020-09-09 at 5.33.18 PM.png

Paglen production image for The Model (Personality), 2020

Screen Shot 2020-09-09 at 5.33.28 PM.png

Trevor Paglen, The Model (Personality), 2020, silver-plated bronze, 18.4 cm × 18.1 cm × 13.7 cm (7-1/4" × 7-1/8" × 5-3/8") © Trevor Paglen

Essays — Bloom: Insights from Trevor Paglen, Sep 10, 2020