The (Impossible) Possible and Perception

In order to think the problem in its entirety, one would have to ask whether the very execution of the algorithm (associated with a given dataset) should not also be apprehended as a place from which forms of uncertainty can emanate. This would make it possible to propose a distribution of the types of uncertainty within the internal anatomy of the digital creative process: algorithmic core, protocol zone and analog periphery.

The pairs of notions considered within a very general framework by Gilles Deleuze – who works here from Bergson –, and problematized in France by Pierre Lévy in the context of the digital, make it possible to launch the reflection at this stage¹. These are the dialectic established between the virtual and the actual, and the one that takes place between the possible and the real². Whereas from the virtual — a dynamic entity that already possesses a “full reality³” according to Deleuze — new forms can be actualized, the possible is already structured; it has already been subjected to an intelligible cutting-up. In this sense, the possible would be too close to the real as what already exists; it would be an “almost real” – to put it in our own words –, from which nothing truly new can arise⁴.

These notions interest us all the more because Pierre Lévy supplies them ready for use within the framework of our research:

[…] if the running of a computer program, purely logical, belongs to the possible/real pair, the interaction between humans and computer systems belongs to the dialectic of the virtual and the actual⁵.

In other words, for Lévy, computer calculation could not take part in the emergence of the new. However, when one begins to dig into the characteristics of these pairs of notions – possible/real, virtual/actual – as they come out of philosophical thought, problems arise that make it difficult to apply them, as they stand, to the problems this research presents.

Indeed, first of all, one should ask whether the notion of the possible that emerges from the reflections of Deleuze and Lévy is relevant for grasping everything that an algorithm, with an associated dataset, is able to calculate: this possible is very close to the real for Deleuze and Lévy; it would “only lack” its realization.

However, returning directly to our object of study, one observation is unavoidable: the number of digital images possible for a given format (2K for example, as on our computer screens), with three color values (Red, Blue, Green) for each of the nearly two million pixels that make up the image, is so vast that it is quite simply impossible to calculate mechanically the totality of the associated images. An overwhelming portion of these “possible” images will certainly never be materialized. Thus, when one considers the spatio-temporal limitations imposed by matter on our machines, the Deleuzian category of the possible – to be grasped, according to the philosopher, as the almost real – in connection with the computing machine becomes problematic, for the complete performance of certain calculations is physically impossible, and very widely so⁶. In fact, no matter how long the algorithm is kept running: it will be able to materialize only a tiny portion of the domain of the possible it addresses.

Yet if the “possible” is not materialized, it cannot become the object of human perception, and it matters little that it is already structured on the intelligible level. This is not a mere detail that distinguishes it from the real, but a fundamental difference. For at the moment of perception, affective, memorial, even psychological components (some of them unconscious) are activated, generally impossible to anticipate. Thus, during the creative process, the plastic relevance of the calculated image, its symbolic radiance and therefore the associated construction of meaning can be settled only in a perceptual relation with the image, that is, after it has actually been calculated and materialized.

Part of this – supposed – incapacity of the possible to generate novelty would stem from the Bergsonian origins of the notion. Indeed, for Bergson, the possible would be a kind of illusion, which emerges when the intellect constructs a halo of possible, past facts associated with a real fact manifesting itself in the present⁷. Yet a possible of such a nature could not describe the domain of possible calculation associated with an algorithm and a dataset — a domain of calculation that is not the illusion born of the defects of our intellect, but that represents a real, albeit machinic, potential.

We therefore prefer not to mobilize the notion of the possible within the framework of this research when it comes to describing the internal dynamic of the digital medium, for this dynamic is regarded by existing theory as an unproductive instance, whereas for this research all the levels of the digital creative process (including the algorithmic core) have a potential for generating novelty.

1. The reflection is to be grasped here as an extension of the considerations that emerged from the structuring works, in the central chapter.↩︎

2. We grasped the interest of these notions during the analysis of the Structurations in the central chapter.↩︎

3. Gilles Deleuze, Différence et répétition. 7e éd. Paris : PUF, 1993, p. 273.↩︎

4. Ibid.↩︎

5. Pierre Lévy, Qu’est-ce que le virtuel? Paris : La Découverte, 1995, p. 15.↩︎

6. We drew this conclusion from the works Every Icon (1997) by John F. Simon, and Every Possible Photography (2013) by Jeffrey Thompson, considered in the central corpus – Structurations subchapter. Indeed, the complete execution of Every Icon, which would calculate all possible icons – that is, monochrome images inscribed in a matrix of a mere 32 by 32 cells – would require several hundred trillion years, which exceeds the age of the universe by several orders of magnitude, even though it is, after all, a very small image.↩︎

7. Henri Bergson, La pensée et le mouvant: essais et conférences. 4e éd. Paris : PUF, 1993, p. 110.↩︎