Abstract
The following theses lay the foundation of a yet-to-be-made research. Asking “how does AI automate Visual Production?” is a question that has to be accompanied by a broader inquiry about how, in History, Visual Production (before and after Art in its modern meaning) has been labour-ified, mechanised and ultimately automated, to furtherly underline the relationships between knowledge, economy and technology, but also to develop or rediscover theoretical categories and concept useful for Art and Image History to react to recent developments in imaginal production.
Thesis 1: The need for a Social History of (automated) Visual Labour. Declaration of Intents
«History of Art is a grandiose History of Labour».
(Antonio Negri)
Following a theoretical front that considers AI under the much larger issue of automation in a broader sense, research aiming to underline an inner logic of labour in contemporary automated visual production, i.e., Generative Artificial Intelligence is greatly needed.
Although “creative” at first glance, automated visual production follows the same logic of labour that characterises any automated system. As Matteo Pasquinelli put it
«The inner code of AI is constituted (…) by the logic of labour and social relations»[1].
The algorithm, the fundamental particle of computational automation responds to the following definition:
«A finite procedure of step-by-step instructions to turn an input into an output, independently of the data, and making the best use of given resources»[2].
In other words, contemporary AI, although more and more complex, still responds to the Rationalität of organised labour, i.e. logical systematisation of behaviours aimed at a predictable result (i.e., patternization of behaviours). This logic, far from being born at the threshold of digitality, constitutes a sociological constant deepening its roots before the so-called computational thinking. Namely, partition of work processes into trivial tasks, social distribution of the latter and related hierarchical structurations (in other words, alienation). This has been the logical basis of the organisation of the Modern State in The Wealth of the Nations by Smith, as well as the engineering foundation of Babbage’s Analytical Engine, the ancestor of computers.
To better understand contemporary automation in visual production a social history of visual labour is greatly needed, in other words how the activity of producing visual artefacts has been divided, organised, labourised and finally automized. Way before computer technologies, image production has always been a matter of labour, but more particularly a business of codification, be it visual, behavioural or social systematisation.
From an art-historical standpoint, the evaluation of formal codes and their rationalisation has been both a concern for theoreticians and the specialist branch of Style Theory and for the highly practical task of drawing up symbological and iconographic compendia such as the books of Emblems and Hyerogliphyca in Medieval and Rinascimental traditions, that sometimes served as instructions on how to build an image.
Furthermore, a social division of tasks, expertise and duties and a relative imposition of hierarchies and commands have characterised a great length of Visual History. In the manuscript Illumination, where a highly hierarchical “production line” was orchestrated to compose codices (pl. for codex); In mural decoration, the same principle has been applied all over the world: to borrow a term from the already cited Pasquinelli, a “social algorithm” of a structured ensemble of hands and minds put in the work to transpose finely arranged forms on a surface to big to handle for a single artist. The list goes on, not without mutation and fundamental shifts, from the coming of Art’s mechanical reproduction[3] to proxy-painting and proxy-Art[4]. The research aims to prove how the history of Automated Visual Production is traceable in a broader History of Visual Labour and that, as other types of labour, sees automation as a programmatic component.
Thesis 2: Criteria of Automated Visual Production are representation of an Inconscio Tecnologico, as Visual General Intellect. Some Basic Assumptions
Given the inner labour nature of automation, both general and visual, the organisation of means of production functions as a magnification of so-called General Intellect[5], namely the crystallised state-of-the-art knowledge on which production is structured and organised. This term of Marxist extraction is often used to identify not obvious relationships between technical, scientific knowledge and socio-economic shaping, physicalised into the logical being of the machine. Critical theories rarely questioned a similar connection of economisation with Visuality as a knowledge area endowed with differentia specifica. Italian photographer (formerly scientist) Franco Vaccari gave a compelling understanding of the relationship between technology and visual knowledge production in the concept of Inconscio Tecnologico (litt. Technological Unconscious), namely the abstract criteria and condition of visual production clotted into photographic technologies[6]. How reality is framed, focused, pictured and thus viewed and perceived is deeply related to how the “labour” of visual production is rationalised, divided and disbursed in the miniature factory that is the Camera.
A similar inconscio could be noted, isolated and analysed in the automated apparatus of Artificial Intelligence, concerning knowledge lato sensu and the particular cognizance that is the Iconic Knowledge[7]. A similar perspective is already taking shape both in practical and theoretical discourse. Leonardo Impett and Franco Moretti experimented with the use of Computer Vision to expand on the Atlas Mnemosyne by fundamental art historian Aby Warburg that is, underlining and analysing formal patterns used to convey specific emotional responses (Pathosformel) and their evolution in art history[8]. Vladan Joler and Matteo Pasquinelli on the other hand, call machine learning a “Nooscope” (from Greek “noos”, or knowledge, and “skopein”: “to examine”) or an “instrument of knowledge magnification”, videlicet a means to probe, to evaluate, to identify, to extract – and finally to capitalise on – various epistemological criteria of knowledge production.
«Instruments of measurement and perception always come with inbuilt aberrations. In the same way that the lenses of microscopes and telescopes are never perfectly curvilinear and smooth, the logical lenses of machine learning embody faults and biases. To understand machine learning and register its impact on society is to study the degree by which social data are diffracted and distorted by these lenses»[9].
For what concerns Automated Visual Production, these criteria are essentially “mimetic”. In a Generative Adversarial Network, the system of algorithms occupied in generating a picture assembles data based on statistics: “form” and “meaning” are associated hinging from psychometric or purely quantitative averages. A generated picture of a dog comes to life based on the “Pathosformel” of a given plethora of (real) dog pictures. A generated abstract painting conveying emotions or sensations (i.e. “labels”) like “anger” or “dynamism” is composed with virtual brushstrokes (actually, visual samples of what is considered a “brushstroke”) “inspired” by real visual data that were psychometrically associated with said emotions.
Thesis 3: {Visual} General Intellect =/= Kunstwollen. A question of method
Formulating a Social History of Automated Visual Labour could tempt one to draw enthusiastically from traditional Art History, seeing categories such as Kunstwollen reified by these new, powerful, statistically derived and arithmetically processed insights. Far from being a ready-to-use “period eye”[10], statistically derived visual metrics archived in the various datasets of Automated Visual Production have to be identified, localised and inquired before becoming an effective historical instrument.
Moreover, considering the specialistic concept of Art, a necessary distinction between (Visual) General Intellect and Kunstwollen has to be made. Kunstwollen, literally “artistic will”, is the concept the art historian Alois Riegl used to explore his Stillfrage (lit. “Problem of Style”)[11], i.e. of what forms and meanings a Visual Culture characterises itself. Riegl proposed an expansion on the speculative range of the Theory of Style, recognizing how an “advanced culture”[12] had to question not only what aims and desires an Art Culture could have, but most importantly how it is conditioned and promoted. Kunstwollen has been only one of the various concepts proposed to give an understanding of the social, cultural and economic drive of the historical unfolding.
For a General, Visual Intellect to be elevated as a Kunstwollen means to responsibilize automated cultural processes. The great difference between Riegl’s concept and our automated inconscio is mainly in a deeper understanding of the verb “to will”.
Measurement biases, technological aberrations, and meta-data are surely a representation of a human – sociologically, geographically and culturally identifiable – rationality, but one of a lower degree than the one that Kant called the “horizon of desirability”. In a Freudian turn of events, the hiatus between Inconscio Tecnologico and Kunstwollen reveals itself comparable to that between the Unconscious and Normative Self. That is between a visceral craving and a programmatic desire.
Summa
In essence, automation proves to be an internal logic of productive action in a broad sense, including visual aspects. A history of visual automation, as a continuation of the History of Labor which has been the history of Art and Style, can and should reveal modes and mechanics inherent in the genealogical evolution of the image in an era where these appear even more obscure and “encrypted”, but above all economically and industrially driven.
Robert Zamboni
Bibliography
Pasquinelli M., The Eye of the Master. A Social History of Artificial Intelligence, Verso Books, Londra e New York 2023.
Benjamin W., The Work of Art in the Age of Mechanical Reproduction, Penguin Books, London 2008.
Galenson D.W., Painting by Proxy: The Conceptual Artist as Manufacturer, in Conceptual Revolutions in Twentieth-Century Art, Cambridge University Press, 2009.
Virno P. , General Intellect, in Historical Materialism 15 (2007), Brill, Leiden.
Pasquinelli M., On the Origins of Marx’s General Intellect, in Radical Philosophy, 2.06 (2019), pp. 43-56.
Vaccari F., Fotografia e Inconscio Tecnologico, Einaudi, Turin 2011.
Boehm G. and Mitchell W.J.T., Pictorial versus Iconic Turn: Two Letters, in Culture, Theory and Critique, 50(2009), pp. 103–121.
Impett L. and Moretti F., Totentanz. Operationalizing Aby Warburg’s Pathosformeln, in Literary Lab Pamphlet 16 (2017), Stanford University Press.
Pasquinelli M. and Joler V., The Nooscope Manifested: An Instrument of Knowledge Extractivism, in AI & Society, 36 (2021), pp. 1263–1280.
Baxandall M., Painting and Experience in fifteenth century Italy: a primer in the social history of pictorial style, Oxford University Press, 1988.
Riegl A., Problems of style: foundations for a history of ornament, Princeton University Press, 1992.
Bianchi Bandinelli R., Kunstwollen, Treccani. Cfr. https://www.treccani.it/enciclopedia/kunstwollen_%28Enciclopedia-dell%27-Arte-Antica%29/ (consultato il 27 ottobre 2024).