As machines are trained to see without human intervention, Trevor Paglen looks at the hidden prejudices and bias inherent in AI with a new Barbican exhibition
In his 2016 essay, Art in the Age of Machine Intelligence, Blaise Aguera y Arcas, the leader of Google’s Seattle AI group and founder of Artists and Machine Intelligence, considers the complex and ever-evolving relationship between art and technological possibilities.
A parallel is drawn between the advent of the photographic medium in the 19th century and the current revolution in machine intelligence, and their profound impact on production and reproduction. He has no doubt that AI will have a transformative affect on our perception of external reality in ways currently unimaginable. As conceptual borders between humans and non-humans become increasingly blurred, he discusses the idea of art generated by ‘hybrid beings’, and how machines extend human thought and imagination.
“In the case of the software,” he writes, “this processing relies on norms and aesthetic judgments on the part of software engineers, so they are also unacknowledged collaborators in the image-making. There’s no such thing as a natural image; perhaps, too, there’s nothing especially artificial about the camera.”
This serves as a useful introduction to the latest body of work from American artist Trevor Paglen, From ‘Apple’ to ‘Anomaly’, on display at The Curve at London’s Barbican Art Gallery from 26 September to 16 February. Paglen has undertaken the 32nd commission for The Curve as part of a year-long programme, entitled Life Rewired, which explores changing aspects of human identity in an age in which we are augmented by instruments and technologies, taking in subjects such as big data and virtual reality.
For this exhibition, Paglen has centred his exploration on the manner in which AI networks have been taught to see the world as a result of being fed vast amounts of visual information by engineers that are sorted into various groupings known as ‘training sets’.
Typically, supervised training sets are collections of known data made up of images, sound and video libraries that train computers to recognise objects or other domain-specific knowledge, such as what humans look like, for example. As the media and computational vision has developed in tandem, machines become autonomous systems that intervene and are coercive in the world, and thus inevitably elements of subjectivity and bias within the research community become particularly apparent within such statistical observations. An investigation into these secret agendas, politics, prejudices and epistemological assumptions, as well as their real-world implications, is precisely what Paglen is intent on probing.
“Machine-seeing-for-machines is a ubiquitous phenomenon,” Paglen has commented, “encompassing everything from facial-recognition systems conducting automated biometric surveillance at airports to department stores intercepting customers’ mobile phone pings to create intricate maps of movements through the aisles. But all this seeing, all of these images, are essentially invisible to human eyes. These images aren’t meant for us; they’re meant to do things in the world; human eyes aren’t in the loop.”
Some 30,000 individually printed photographs are to be installed across the entire surface of the curved wall in the Barbican gallery, forming a complex mosaic organised according to more than 200 categories selected by the artist. While these are labelled so that visitors can identify the respective classifications, no further textual explanation as to the reasons for the choices is given. To understand the lineage of this body of work involves a brief outline of Paglen’s source for the images – ImageNet, a publicly available dataset consisting of annotated photographs intended for computer vision research and understanding algorithms.
There are more than 14 million images in the dataset, with over 21,000 categories, and one million that have bounding box annotations that identify objects within the images. These were culled directly from the internet by academics made up from a consortium of American universities, including Princeton and Stanford, and, curiously, the ImageNet project does not contain the copyright for the material.
The appeal to Paglen, however, lies in the politics and practices of categorisation. For the most part, these are benign, as is the case with, say, ‘strawberry’ or ‘orange’, while other classifications take on more untoward implications, such as those filed under ‘debtors’, ‘alcoholics’ or ‘unusual person’.
Evidently, the act of programming is also one of making judgments. And although the images may elude artistic signature – since it is the language that speaks and not the authors within this empirical mass of images – the groupings nonetheless reflect an inherent lack of impartiality towards their subjects. As Aguera y Arcas and countless others before have reminded us, technology is never neutral. These thoughts press harder even when we consider the significance of the two specific categories used with the exhibition’s title; apple and anomaly – one is a strict noun, the other relational, thus its connotations are potentially open to misdirection.
As a proposition, From ‘Apple’ to ‘Anomaly’ invites the viewer to consider that the world of images has grown distanced from human eyes as machines have been trained to see without us. Paglen often refers to this new state of machine-to-machine image-making as “invisible images”, in light of the fact that this form of vision is “inherently inaccessible to the human eyes”. He has also on occasion posited that we are perhaps operating within a surrealist moment for images, similar to the semiotics that come to bear in René Magritte’s Ceci n’est pas une pomme. So it’s no surprise that the famed painting serves as the subject in one of Paglen’s individual works, The Treachery of Object Recognition.
Throughout his career, Paglen has developed a long-standing interest around issues of surveillance, CIA black sites, drone warfare, the essence and apparatus of America’s security systems, and much more. His is a practice broadly underpinned by an investigation into the relationship between vision, power and technology.
Paglen’s work has been widely exhibited at institutions internationally, ranging from Tate Modern and The Met to biennials in Taipei and Istanbul. Among many awards and accolades, he has been the recipient of the Deutsche Börse Photography Foundation Prize 2016 and the prestigious McCarthy Fellowship in 2017.
Now based in Berlin, he remains one of the most urgent chroniclers of our times, highlighting the forces that lay beyond what is immediately evident. This exhibition is testament to a highly original artist, always on the move, always enquiring, re-inscribing what it means to learn to see, all the while keeping a critical and more responsible relationship to the world, to what we understand of the world of images.
From ‘Apple’ to ‘Anomaly’ by Trevor Paglen is on show at Barbican’s The Curve gallery in London from 26 September 2019 until 16 February 2020