“...
there is no difference inprinciple between a universe,
a religion, and a hoax.
All involve an engineering of
manifestation, or practical
fiction, that is ultimately
unworthy of belief. Nothing
is true, because everything
is under production.”
― CCRU: Writings 1997 - 2003
︎
About us
Text
Shop
Contact Us
Submissions
Solarisation: Accelerating (Machine) Vision into Darkness
What does it mean to ‘see’? What does it mean to represent reality? And how can machine vision models guide us towards a dark vision of reality which moves beyond the vision of enlightenment? I will speculate on these questions in this text. Through the example of solarisation, I will explore to what extent computer vision models can throw us into an epistemology of darkness.
The Epistemology of Light
To contemplate the sun would be the definitive confirmation of enlightenment.1
In The Thirst for Annihilation, Nick Land showed how European philosophy has been obsessed with relating the sun to the valorisation of truth.2 Whereas Plato sang of 'the glory of one sun', Kant insisted on the light of pure reason and Heidegger talked about the Lichtung, or clearing, of Being. They seemed to share the conviction that the sun allows for a pure illumination simultaneous with the increase of truth. Seeing, here is seen as an intentional movement, the gaze of the eye, impressed upon the outside. An organic model of endogenesis where the intentionally created visions of reality, social repressions, are what generate and drive systems of knowledge towards ultimate enlightenment.
Light and Machine Perception
We find this sentiment also in current conceptions about machine perception. Contemporary machine perception often operates through deep learning computer vision models, for example in object detection. Here, the goal is to detect objects in an image or video and classify them into categories.3 However, the categories of the objects to be detected are defined before the training process and the accuracy of the model is determined based on the correct categorisation of the recognised objects in those categories. It learns such a correct categorisation through organical error reduction. For each iteration, the errors made by the model are back propagated through the network, thus updating its weights and, consequently, its learned representations.4
We see this for example in the image below, depicting a classic plot of the training progress of an object detection model.5
This image depicts training and test errors (y axis) of two different models (the red line and the yellow line) through time (x axis), where time is defined as the number of training iterations. As such, these plots clearly show that typical machine perception arrives at a notion of truth through organical error reduction. As such, the models' recognition process of reality can be said to be determined by this outward movement of structuring reality by means of technological error reduction which characterises the epistemology of light.
The fact that contemporary computer vision models follow this epistemology is even more highlighted due to the fact that they, by using images or videos as training data, need light in order to be able to recognise anything at all. Dark object detection is a notoriously hard problem in computer vision.6 Due to the lack of detail in an image, the object boundaries cannot correctly be recognised. This can be improved through an increase of brightness in the image, because the increase of brightness can improve detail. As such, the epistemology of computer vision models seems to adhere, at least partially, to the statement that 'more sunlight automatically translates into more reality'.7
The Black Sun
However, following Lands categorisation of Bataille, this reality of the sun is not the only possible reality:
"Mixed with this nourishing radiance, as its very heart, is the other sun, the deeper one, dark and contagious, provoking a howl from Bataille: ‘the sun is black’".8
The sun is not only a harbinger of truth and knowledge, but also one of malediction. It infects whoever’s eyes fall upon it. It burns, wherever its rays touch. It provides us not only with illumination, but also with hallucination.
We experience this, for example, when we look directly into the sun: "Gazing into the golden rage of the sun shreds vision into scraps of light and darkness. A white sun is congealed from patches of light, floating ephemerally at the edge of blindness".9 The light of the sun fractures our sight; wounds our retina, providing us with a schizoid view of reality completely inhibiting any cognising of the world.
We find another example in the description of 'solarisation' from the book Revolutionary Demonology. Solarisation is the process in which an image becomes overexposed to light during the development process. Consequently, this overexposure of the image produces the effect of ``t0nal inversion``; black becomes white, white becomes black, leading us to experience the sun as black and the sky as white.10 As can be seen in Minor White’s Black Sun below.
Importantly, from the perspective of the epistemology of light, the Black Sun created by solarisation represents a wrongful representation of a real phenomenon, made possible only through the lens of a technological artefact (the image). The sun is shown to be black, whereas, in reality, it is white.
Yet, from the perspective of Revolutionary Demonology, this wrongful representation is not one to be discarded, but rather to be affirmed as a vector of change. It is only through the mediating filter of the camera, and the repurposing of it qua tempering with it, that the de-illuminated reality of the sun reveals itself: “the cosmic power of the sun explode[s]... the unknowable emerge[s] from the shadow zone in which the veil of phenomenal reality had confined them.”11
Although these examples illustrate the idea of the black sun, they do not do justice to its full scope. For the real meaning of the darkness in the black sun is not just a fractured retina, nor the sun depicted as a black spot, it is the realisation that we have been hallucinating the entire time and that, in fact, hallucination is all there is to knowledge. The identification of knowledge with illumination is a hallucination. That what we see; the way we see it; does not approximate the full breadth of reality. Rather, the vision of light is a 'fractured cunning of reason' made possible by the abstractionist logic of technocapitalism in its process of illuminating itself to itself. After all; 'the ultimate capital is the sun',12 Serres writes. The pursuit of 'truth', the accumulation of solar energy to bring the human race enlightenment, serves the capitalist tendency towards abstraction. It obfuscates the reality of nature as unplanned synthesis;13 a schizophrenic, “cyberpositive mutation, at war with the complex of organic judgement”.14 A war, we might say, with the recursive process of organic error reduction.
As such, ``see ing in the light is blindness``. The real power which the darkness of the black sun exerts upon reality is this realisation. Fractured seeing is only undesirable because it leads to a schizoid experience of reality: the effect of solarisation is only an error when we have a specific goal in mind. Similarly, wrongful classifications by computer vision models are only erroneous within the teleological framework of representing reality as a true vision of light.
Dark Machine Vision
Then, how do we get out of the light into darkness? Should we reject the hallucinatory
capacities of the sun, the erroneous process of solarisation, in favour of a truthful way of seeing? Or should we, perhaps, integrate error reduction organically into our vision of reality, like computer vision models do?
No, these errors are not something to be rejected. It is only through these instances of
temporary intensification of the sunlight, error intensification by computer vision models, the rapid increase of the hallucinatory capacity up to a breaking point, that we can come to the conclusion of the black sun. After all, “the indelicate philosophical instrument of ‘presence’ has atrophied our eyes to such an extent that the dense materiality of light scarcely impinges on our intelligence.”15 Our eyes are scattered and so is the relation between light and truth; illumination and enlightenment. We can not go back: we can not decelerate.16
Instead, the only way forward (or rather, non-directional into the depths of darkness) is to
intensify these technological errors, accelerate them up to the point that the hallucinatory capacity of the light sun flips into the darkness of the unknown. An example of this could be to train machine vision models on the hallucinations of other models. For example, people have started using Luma’s recent model “Dream Machine 1.5”, in order to create weird gymnastics videos where bodies morph into other bodies, fall through the bar and fly in the air without seemingly ordered adherence to physical laws.17
From the point of view of the epistemology of light, these scenes can be seen as errors on the path to correct video creation; hallucinations of reality to be discarded. However, another approach would be to train vision models on videos like this, as to further disalign their object representations with how we, humans, represent objects. Such a practice could throw us into darkness, because it goes against the categorical material structuring. No categories suffice to categorise the formless morphing objects. As such, machine perception could be deployed in the dark and could contribute to an acceleration of our view of vision out of an epistemology of light, into an epistemology of darkness.
Bibliography
1.Nick Land, The Thirst for Annihilation, 1992, p.20. Auslander, P. (2011) Liveness: Performance in a mediatized culture. London: Routledge.
2.Ibid.
3.Joanna Zylinska, “Machine Perception” in Chimeras: Inventory of Synthetic Cognition, 2022, p.27-p.29.
4.Yuk Hui, Recursivity and Contingency, 2019.
5.Kaiming He, et al. Deep Residual Learning for Image Recognition, url: https://arxiv.org/abs/1512.03385.
6.Xiangchen Yin, et al. "PE-YOLO: Pyramid Enhancement Network for Dark Object Detection", in Artificial Neural Networks and Machine Learning – ICANN, 2023.
7.Gruppo Di Nun, Revolutionary Demonology, 2023, p. 209.
8.Ibid
9.The Thirst For Annihilation, p.20.
10.Revolutionary Demonology, p.207-p.209.
11.Ibid, p.210.
12.Michel Serres, Le Parasite (Paris: Grass, 1980). English translation: The Parasite (Baltimore: Johns Hopkins University Press, 1982), p.173.
13.Nick Land, "Circuitries" in The Accelerationist Reader, 2024, p. 270.
14.Ibid., p.271.
15.Nick Land, The Thirst For Annihilation, p.20.
16.Robin MacKay, Armen Avanessian, 2014. The Accelerationist Reader Urbanomic
17.@werners_ai_art, https://www.instagram.com/p/C-60sa8Ru3t/?hl=en .
@ivarfrisch