Taxonomy: T240412-32.8.416/T240412-32.4.412/T240412-32.6.414/T240412-32.2.410/T240412-32.14.422 (2024)

GESTALTOLOGY

by DiPisaStasinski, 2020-

Research project in collaboration with:
Alessandra Di Pisa, artist
Robert Stasinski, artist
Gabriel Eilertsen, associate professor, Machine learning, LiU
Jonas Unger, professor, Computer Graphics and Image Processing, LiU
Rike Stelkens, associate professor, Evolutionary Biology, SU
Jonas Löwgren, professor, Interaction and information design, LiU
Apostolia Tsirikoglou, PhD, LiU (2020-2022)
Per Larsson, engineer (2022-2023)
Cecila Åsberg, professor, Gender, nature, culture and posthumanities, LiU (2024-)

Artificial intelligence, evolutionary biology, and branches of visual regimes meld together in the project Gestaltology, where AI is a fundamental tool and the subject of critical exploration. In this ongoing work, the artists have partnered with researchers in AI, engineering, design, and biology to form a unit of exploration in science and artistic research. The mode of exploration is tinkering with possibilities of new technologies, challenging the anthropocentric visual tradition, and finding alleys to explore alternative paths to develop aesthetic processes in technological advancements.

GESTALTOLOGY relates to the concept of The New Aesthetic, which supports the shift from machines being merely an extension of human creativity to a more autonomous form of creation.

Our latest iteration is ARCANA - an artificial phenotype digitally linked to its environment and constant struggle, learning to create a physical Gestalt on paper. ARCANA shifts its aesthetic preferences through its bodily movement. The exhibition consists of a spatial installation with one of the artificial organisms actively engaged in its environment through physical and algorithmic processes. The installation showcases generational lines of drawings executed by the organism ARCANA, which, in turn, have evolved the organism to the next generation in an ongoing loop.

The term Arcanum once entailed the practical secrets that the alchemists of the 20th century tried to discover through the active making and manipulation of materials through meta-scientific tinkering.

Enchanted threads of visual Archana - GESTALTOLOGY ENCODED

Installation, Visualiseringscenter Norrköping, 2023

(LED light, acrylic, paper, video projection, electromagnets, wood, metal, rubber, plastics, cardboard, spray paint, visual sensors, Raspberry Pi)

Photo: DiPisaStasinski

Arcana - perched atop a vertical synthetic vegetation - appears as a seemingly modest little object, consisting of a tangle of intricate thin metallic threads. But illuminated with a strong light from above, a spark of life seems to ignite, and the object undergoes a metamorphosis. With its tentacle-like limbs, Arcana begins to tentatively explore its surroundings.

Arcana is the small organism in DiPisaStasinski's work Enchanted Threads of Visual Arcana at the Visualization Center, 2023. Arcana is also a concept with a long and sprawling history, from a Latin expression for mystique, to the early 20th-century alchemical visions among European chemistry amateurs, but also the name of the Tarot cards' major and minor arcana, which represent Jungian archetypes of the collective unconscious.

However, Arcana at the Visualization Center is more than an artificial organism; it is also an environment - a garish non-organic sci-fi-like vegetation sprung from low-tech material, housing high technology, with which the organism operates in total symbiosis. Likewise, Arcana is a process, depicted through hundreds of small images, sometimes produced as a sequence of mathematical cuts, other times as development lines in an ecology branching in various directions, and it is in the relations between these that the concept and forms gain their relevance, much like the Tarot cards through their ritualistic random placement, create a seemingly meaningful order.

The small creature leaves traces behind, imprints that form color clusters on the surface, similar to the paintings that encircle the installation. Suddenly, it stops, the image it created disappears, is stored in memory and after a moment of reflection, Arcana begins anew. The wall paintings can thus be understood as just memory fragments, parts of the creature's history and shape, through which it has evolved.

Taxonomy: T221123-27.1.307/T221123-27.2.308/T221123-27.3.309/T221123-27.4.310 (2022)

Evolutionary Traces of Arcana (2022)

Malmstens Gallery, 2023

(Video, acrylic, paper, magnets, clay, tape)

Photo: DiPisaStasinski

Commercial models vs. autonomous organisms

by Sebastian Rozenberg

The ‘intersection of art and technology’ is a worn out cliché by this point. Nonetheless, art is often the site for negotiations around AI, and artworks are more often than not used as illustrations in research and critique. “AI” seems presently to evoke an almost mystical force, ascribed various powers or threatening capabilities (Crary, 2023), potentially catastrophic to both planetary welfare and human creativity alike. Others seek to mobilise against and educate on the hidden extraction of data and exploitation of labour of models, systems, and platforms that are variously given the name Artificial Intelligence (Aradau & Bunz, 2022). A third group of accelerationists and techno-optimists wish to speed up technological advancement and welcome the idea of an artificial general intelligence governing and steering humanity right. Art that incorporates or works with AI either conceptually or concretely, often actualise specific positions and relations between tools and collaborators, between human and non-human systems or organisms.[1] The trouble starts when trying to say whether one determines the other, art or technology?

As I see it there is no easily determined teleological movement from one to the other. If the debate is set between technology determining the art or the reverse, the argument is already muddled. Both appear as crystallisations of human activity (Simondon, 2017). Instead, good arguments can be made for an indeterminate relation between technology and art, as Pierre Francastel claimed 70 years ago: “Our epoch has the art which it merits and this art is far from being inconsistent with technique” (Francastel, 1953). But, although the history of computer art and information aesthetics is intrinsic to contemporary concerns regarding art and computation, I would categorise very little if any of what is produced by mainstream AI models as art. Most images and media created by generative AI platforms is more accurately described as statistical renders or algorithmic remediations. Driven by a market logic of effectivization and optimisations, a teleological drive for growth. The popular commercial platform models for text-2-image generation, such as DALL-E, Stable Diffusion and Midjourney generate what Hito Steyerl calls mean images. That is statistical renders weighted towards the median, the generic and stereotype laden median of the training data set. Avoiding the hyped up and obfuscating language of AI, these image generators can be more accurately described as automated pattern recognition models. Critique of computational creativity and AI is often not so much a critique of the model or specific software, but rather a critique of the violence of contemporary capitalist systems of algorithmic governance as well as its economies of attention. Just as late capitalism restricts the possibilities of imagining and thinking productively of solutions to planetary scale crises (Crary, 2022), the platforms and the capitalist relations they entail both restrict and form the human-machine relation.

***

Arcana can be situated as a specific form of resistance to this text-2-image models and similar AI technology, an alternative to the quantitative screen-based logic of platforms and the aesthetics it endorses, one based purely on data and attention. The little metal body or body part of the sculpture computation system seems to be struggling against invisible restrictions. This is a stark contrast to the smooth ease of commercial generative machine learning models, operating within the determining horizon of a training data set. As an evolutionary program – a phenotype allowed to develop its own quirks and retain code that is normally deemed faulty – Arcana acts ostensibly outside of human culture, creating its own input. Seen on a scale from open, distributed, and networked to closed, isolated, and contained, Arcana gives little of its output or processes to the viewer. Neither is it plugged in to the global network that looms both large and abstract among both technooptimists and doomsayers.

Arcana is an evolutionary algorithm, a piece of code that evolves and develops, meaning the main part of the work is operational, has no representative form for human consumption. The sculptural installation that anchors the work, appears almost as a concession to the human lust for representation and visible concretisations of processes. Text-2-image models by comparison are almost only representation, a tool for producing representations from large ensembles of images. Perhaps it is more constructive to understand them not as tools or self-directed creative machines, but as gimmicks, as ”both a wonder and a trick,” in the sense elaborated by writer Sianne Ngai. The gimmick “is a form we marvel at and distrust, admire and disdain.” Our feelings for a gimmick slip between good and bad, it is “enchanting and repulsive at once” (Ngai, 2020), eliciting both admiration and fright. We’re scared the AI model will take away what it means to be human, or simply replace the work of artist, but impressed by the images it produces. Central to this is the understanding of the gimmick as a labour-saving device – it appears to work too well and too little at the same time, both saving time and stealing jobs. In its basic definition the gimmick is also a device for capturing attention. If in a surveillance and platform capitalist society data extraction is the (perhaps less and less) hidden centre, the economy and capture of attention is the interface and front end. The opaque nature of Arcana on the other hand is the opposite of a gimmick, not a wonder or a trick. It appears aimed at preserving the inefficient and contingent, not an effort to automate human thinking or creativity. In comparison to practically all commercial AI models, Arcana only extracts and abstracts from its own previous labour, not from human labour. While the gimmick and its labour-saving promise goes hand in hand with acceleration and optimization, Arcana appears slow and inaccessible, intentionally, and intrinsically difficult to explain and anthropomorphise. But we want to ascribe technology human properties, speak of it in human terms. Philosopher of technology Ernst Kapp identified this tendency in 1877 already, arguing that human beings lack “any appropriate expression for non-human conditions.” We categorise apparent similarities of technology in human terms but should never conflate our image of the thing with its true workings (Kapp, 2018). We want technologies to mirror us, we see the human in the machine.

But this is an anthropomorphic error. Aesthetics for machines, the sensory knowledge of machines, is inaccessible and autonomous, “beyond the limits of human knowledge” (Fazi, 2020). If the machine has sensibility, it is a sensibility of itself performing computation (Fazi, 2019). A freedom from human knowledge is a form of autonomy from all types of human abstraction, which includes art and philosophy and perception. In a certain sense this is already happening with algorithmically calculated culture, where the consumer is subject to certain standardised ways of consuming and experiencing, and is less and less a participant. Bernard Stiegler calls this the proletarianization of sensibility. He sees the use of digital network technologies by capitalist culture industries as a proletarianization of the spectator and participant in culture, meaning a downward mobility where citizens, spectators, users are no longer actively part of the creation of symbols and artefacts, but rather consumers only (Stiegler, 2014). And, we could add, unwitting producers of data for the models, that are ostensibly creative without humans, but still create for a human audience – at least until consumption and experience itself is also algorithmically automated. While this is perhaps an example of a tragic view of culture and society, far to one sided and pessimistic, it is nevertheless true that what was once a spectator and participant is more and more just a consumer today.

***

Can the autonomy of art and the automation of thinking find common ground? This question is repeated in both research papers and popular articles to almost comedic effect. Can machines create art? Can computers create art? Can a machine think anything new? As the large majority of the systems discussed in this way are part of large platforms or ostensible non-profits (OpenAI is mostly for-profit and massively funded by Microsoft), the question must be viewed as part of the obfuscation and shifting of perspective that is currently taking attention away from harmful aspects of AI systems, forces interested in the reproduction and sustaining of data colonialism and platform capitalism. For any artwork to be autonomous and have the potential for harbouring critique, it must be created at least partly outside of the parameters of platforms and corporations.

The question human and machine, organic thinking and inorganic reason can be conceptualized as a question of anthropocentrism in opposition to an originary technicity of the human and all organisms. The technical as intruding and controlling or the technical as constitutive for human thinking and remembering. Of course, this opposition need not be so clear cut, current theories argue for concepts of cocreation with machines and models, or on a more conceptual and axiological level, a split vision, sharing the perception with machines, allowing the computational to shape our thinking. What about sharing creativity and art with machines? This would not be problematic in itself were it not for the platformed nature of the majority of the models. Our aesthetic judgement of AI must take its commercial and harmful practices into account. In terms of what Arcana actually does – as the staging of a process the viewer is invited to observe – it is an evolutionary image model, a phenotype developing its ‘own’ aesthetics and image consciousness. Maybe not a tool for making art then, but still a mechanism for observing certain aspects of rational and creative thinking. Perhaps our own?

This tension is also reason to reflect on autonomy and automation, as a question of different types of self-determination. If all computation is a process of determining the indeterminate (Fazi, 2019), of discretising reality, this determining is mirrored in the self-determining of a model which produces its own training data, and again in an artwork that asserts its self-determination through a disavowal of mediation and interfacing, and a refusal of simplicity and statistically streamlined image renders. Luciana Parisi argues that ascribing (or giving) reason to machines means not a liberation from the “responsibility of decisionmaking and self-determining judgment,” but rather that AI as an ‘alien subject’ exposes the capacity of a programmed medium to develop a transcendental dimension (Parisi, 2019). More simply put, this type of system generates a type of subjectivity that is alien to human understanding.

Any artwork or image created through and within a platform model is forever caught within a corporate system of profit. Arcana on the other hand, isn’t modelled on human thought or human creativity exactly, it isn’t anthroponormative. But avoiding all anthropomorphism, apart from that instilled in the structure of all algorithms as socially conditioned practices, the machine must turn completely inward, presenting only the indecipherable and almost parodically machinic movements of the metal operator on the light table. The actual process the work performs is aimed at image generation, but unlike a text-2-image model, it escapes all kinds of aesthetic measure, leaving room for neither human nor outside algorithmic judgement. In the indeterminacy between different modes of function and existence, between evolutionary algorithms and statistical renderings, between closed systems and globally spanning networks, between recognizably human images and a sensibility of machines, there lies a possibility of art that doesn’t shy away from the complexity of human and more-than-human interactions. Any autonomy of this type of non-human organism is only possible through an openness towards that which will partly remain closed to human viewers, an openness that must amount to care for a subject, whether real or imagined, that is alien to us.

This is also the constitutive tension as well as the potential critique of algorithmic culture Arcana harbours - the system creates images but withholds them from circulation as well as rating and judgement. While the human spectators aren’t subjected to an algorithmic personalization or governing of cultural consumption, neither are we exactly allowed to participate in or given access to the resulting images. The question remains, are we part of the cultural production of symbols and images, or passive users? Here we are neither. That’s a refreshing change.

[1] What Yuk Hui calls the organising inorganic, or James Bridle more concretely terms more-than-human.

Photo: DiPisaStasinski

Sketches by DiPisaStasinski