(Self-) Portraits of An Algorithm

Algorithms are the operational oligarchs in today’s complex societies, as they control, supervise, initiate, terminate, and regulate an extraordinary amount of human activities. As a consequence, algorithms ought to represent an intriguing subject of photography. Unfortunately, because of their immateriality, algorithms fail to deliver that critical photographic requirement of being able to reflect photons toward a sensing surface. And yet – if we just consider those algorithms that dominate photography – we “see” many of them all the time: they are busy restructuring the real, morphing our perception into outrage or happiness or anger, challenging norms or tactically accepting hypocritical (because they never criticize, in their pure political correctness) truces. Their relentless and suffocating presence has made some of us rethink the whole meaning of photography [1].

And yet, important actors they are. Is it then possible to take a portrait of an algorithm? What does it mean, actually? Algorithms could be portrayed through a formal mathematical language. Accuracy of representation and flawless formality would be guaranteed but intelligibility would be mediocre at best, and there would be no empathy between viewer and subject. More to the point, a mathematical representation would prevent us from appreciating the inner self of the algorithm, something similar to claiming that an MRI of the brain depicts the conscience of an individual. Finally, we are much more interested in intelligibility than in accuracy, and the grammar of choice here is neither mathematical nor textual, but visual.


But what about the immateriality issue noted above, can we overcome what appears to be a trivial but insurmountable roadblock? Well, any algorithm can be portrayed via mediation or, better, via transference.

Transference happens between the algorithm and the material and this can be captured on camera. Transference happens between the artist and the material too, but –­ as we shall conclude at the end of this paper – the artist here is the algorithm, certainly not the photographer, who is a necessary but irrelevant presence. Kuspit says that “Transference is in effect the basic creative act – one might say necessary emotional work – that makes the material into art” [2]. Here transference makes the invisible become visible, the immaterial become capturable, the algorithm become the artist. Transference is what is captured, in fact, not the algorithm per se. Particle physics operates in similar fashion all the time: the portrait of a subatomic particle is often the picture of what the particle has left in its traveling through detectors. And this portrait, in spite of its not showing the particle at all, is deemed to carry so much significance as to represent the proof of that very particle’s existence. In a field that could not be further away from particle physics, i.e., paleontology, the fossil print left by an extinct animal is proof of the past existence of that animal.

Van Lier writes: “… a photograph is not merely a blend of reality and the real. It is a phenomenon where what is represented of reality comes to us across the frame of the real. (Emphasis his.) [3] The real, in Van Lier’s words, “is all that is before, after and underneath reality, it is all that is not yet domesticated by our technical, scientific , and social relations.” [3] If we subscribe to these definitions – and we do – we are at great difficulty to see them applied in this project, however. Distinguishing between real and reality in the context of portraying a completely elusive, opaque, and immaterial subject as an algorithm is not possible: real and reality have lost their meaning. The real – the high energy particle traveling through detectors or the algorithm cannot be domesticated by whatever relation we can come up with, because of that complete elusiveness we mentioned before. The reality that should supposedly come across the frame of the real has lost their meaning too, if not because there is no frame! While this reasoning seems to have a logical and self-consistent strength, the nihilistic, all-is-lost conclusion does not sound right. We shall resolve this intellectual conundrum in what follows.

Finally, there is also a time component. What is captured is a before and a now, the difference between the two being the transference, i.e., the portrait itself. However, while the now is in front of the viewer of the portrait, the before shall be in his or her memory, and ought to be universal enough (universal in the sense of being shared by all humans, regardless of geography, culture, or traditions) as to allow the viewer to appreciate the transference effect, i.e., to “see” the algorithm. This universality is the sole territory of very familiar objects. What better choice than to choose simple fruits and vegetables? Their shapes, colors, and textures are immediately recognizable by any adult.

Our project presents twenty portraits of the Singular Value Decomposition, also known through its acronym: SVD. The SVD is a well-known linear algebra algorithm used in many real-life applications, from signal processing to material science, from internet searches to economics. It is without a doubt one of those operational oligarchs that control our daily lives.

Taking the portrait of an algorithm challenges some norms in both the aesthetics of photography and its philosophy. To put this in a conceptual framework that allows us to reason about the project, let’s assume the classical precursors of any photograph: the operator, the apparatus, and the subject. They conspire to produce a photograph; their cross-relationships have been an important subject of investigation for philosophers and other intellectuals.

Interpreting the role of each of these three precursors in the context of this project immediately presents us with a number of puzzling questions. This was to be expected: the real-reality trap we discussed above had given us an early warning that something was not well understood.

The operator. Derrida talks about photographic act [4] as opposed to photograph. Flusser talks about photographic gesture [5]. Our intuition, however, tells us that there is no act or gesture in this case, at least of any consequential artistic or semantic value. Moreover, many methodologies call for putting the photographer‘s (operator’s) personal information at the center of the reading of his or her body or work. Here the operator is largely irrelevant; he or she has to deposit a piece of fruit on a light table to be photographed but this is something a robotic arm could accomplished as effectively. We do not see the gesture. We do not see the act. Finally, knowing the biography of the operator is of no use in reading the photograph. These fair statements conspire to articulate “the” question that has been around as an undercurrent in the discussion so far: is the operator still the artist? We postulate here that the artist is not the operator but the algorithm itself.

The apparatus. The apparatus produces what Flusser calls technical images [5] – to distinguish them from those produced by painting or through other activities – and for him they have a magical effect. It is magic that “conjures tricks with abstractions”. [5] Not only are we unquestionably dealing with technical images in this project, but his two words, i.e., trick and abstraction, apply to this project in many surprising and fascinating ways. Inside the apparatus the indexes take form, and where the powerful programs ­that supervise the production of the photograph operate. This is where the discourse starts drifting away, if applied to our project: yes, a camera was used for these portraits and the programs were “acting” upon the sensing surface hit by the photons. But the apparatus does not change the meaning of the world, as Flussers claims (correctly) it does: it simply supervises the mechanics, once we realize that that these photographs record a transference, and not reality through the real. The photograph was not even controlled by the algorithm being portrayed, because the algorithm appears to be the scenery (the subject). So, what does the apparatus do in this case and how can it be defined?

The scenery. The classical theory defines it as what is in front of the lens. This is not the case here, as the photograph does not depict what was in front of the lens. In fact, the viewer looking at the photograph does not see a fennel or an apple. Given the exceptional familiarity that viewers have with the objects (and they were indeed chosen because of this), the former instinctively see the difference from a ‘normal’ fennel or apple. Viewers see the transference, i.e., they see the algorithm. This is the scenery, therefore, and it is not what is front of the lens. Much like the classical understanding of operator and apparatus came into question in this project, the scenery too is not what is normally understood, although it carries some intriguing resemblance to Sontag’s trace [6], while it lives very far apart from Barthes’ creator of the Referent [7].

Let’s make something very clear, though: the difficulty we face in organizing these portraits from an epistemological point of view is not because there is ex-post data manipulation carried out by the SVD. Millions if not billions of photographs are processed by algorithms of all sorts and still naturally fall into the operator-apparatus-scenery model and can be interpreted according to the latter. The issue here is that what is correctly claimed to be the focus of the photograph – i.e., the algorithm – is actually part scenery, part operator, part apparatus, and part photograph.

Stephen Shore wrote: “A photographer standing before houses and streets and people and trees and artifacts of a culture imposes an order on the scene – simplifies the jumble by giving it structure.” [8] This powerful sentence is a wonderful description of the photographer’s role. The problem is that it is the algorithm that imposes an order on the scene in this project, certainly not who cocked and pressed the shutter. This is the reason for the title “(Self)-Portraits Of An Algorithm.” We may stretch the logical conclusions somewhat, but it is not an attention-grabbing provocation.



If all of the above intellectual difficulties were to be dealt with through a novel understanding of operator, apparatus and scenery/subject, we could read the SVD portraits. In fact, there is a way to make all the exceptions and discontinuities in our reasoning lose their toxicity within the three precursors, but it requires a major shift in perspective. The need to take their toxicity away is not to be able to have an instrument to read these photographs we accept the notion that many photographs cannot be read and we feel just fine about it. The toxicity needs to be taken away because this major change in perspective is the intellectually honest path to pursue. The fact that this re-establishes congruency within the operator-apparatus-subject triad and hence makes reading the photographs possible is a welcomed by-product, nothing more.

Our intuition proceeds as follows: the algorithm is Derrida’s photographic act, it is in part the subject, the photograph, and the operator. Then we can conclude that the algorithm is the artist. These are self-portraits of an artist: the SVD algorithm. 

If we are willing to jump over this sky-high intellectual fence all problems mentioned above cease to exist and our precursors regain their logical consistency. And the photographs in this project can be read. Indeed:

  • The Operator is the Algorithm and it is a self-portrait.
  • The apparatus is the set of programs controlling the camera, as the standard definition stated. The algorithm being photographed is no longer part of the apparatus, being the operator and the scenery.
  • The scenery is the transference, which is the algorithm, but also the artist. It is a self-portrait, after all. This is a bit less straightforward than the previous two. The definition of the scenery as “what is in front of the lens” applies in fact to “conventional” photography, i.e., photography that is neither artificial life nor computer generated imaging. These self-portraits fall into a grey area, i.e., neither real not artificial, they are the photographic equivalent of the Borg of the Star Trek saga. However, the photograph depicts the transference and Van Lier’s distinction between “real” and “reality” applies verbatim: the real is the vegetable in its disorganized being; reality is the capturing of the transference – the SVD portrait ­– i.e., the ‘real” after being structured by Shore’s photographer: the algorithm.

This project has opened up a can of worms that goes from aesthetics to philosophy, from perception to epistemology. We have no claim to have addressed the problem and made sense of its intricacies. With lots of fallacies we have just scratched the surface. We agree with Van Lier when he says that “…the photograph frustrates nearly every property of perception” [3] and this project, we believe, has moved this frustration to a new level. We are sure of one thing, however: the unstoppable invasion (generally helpful, at times perverse) of algorithms in photography will force a profound rethinking of some fundamental concepts in photographic art that so far have been assumed as self-evident truths. The questions that derive from this rethinking are scaring in their complexity: Who is the artist? Worse even, is there still an artist? What does a photograph truly represent? Does it still make sense to reason about indeces? Is the viewer the consumer or the producer? Are we making room for algorithms in Art as we have already done in business, in health-care, in transportation, in every day life, allowing them to move from a subservient to a (only in part, for now) dominant position?

 Flusser wrote: “Prior to the Industrial Revolution the human being was surrounded by tools, afterwards the machine was surrounded by human beings.” Is the next step to get rid of the human beings and leave only machines (i.e., the algorithms, in Photography)?

To put it bluntly: Is this liberal attitude of ours towards allowing an ever increasing presence of algorithms in Photography tantamount to welcoming a Trojan horse into our collective artistic psyche?


[1] Fred Ritchin, After Photography, W.W. Norton & Co., 2009.

[2] Donald Kuspit, The End Of Art, Cambridge University Press, 2004.

[3] Henri Van Lier, Philosophy of Photography, Leuven University Press, 2007.

[4] Jacques Derrida, Copy, Archive, Signature: A Conversation on Photography, Stanford University Press, 2010.

[5] Vilém Flusser, Towards A Philosophy Of Photography, Reaktion Books, 1983.

[6] Susan Sontag, On Photography, Farrar, Strauss and Giroux, 1973.

[7] Roland Barthes, Camera Lucida, Hill and Wang, NewYork, 1981.

[8] Stephen Shore, The Nature Of Photographs, Phaidon, 2007.


Leave a Reply