Repairing Rationality: Max Bense and the Automation of Literature in Post-War Germany

Introduction: Jetzt / Now

JETZT, jetzt und erst jetzt, jetzt und nur jetzt, jetzt und doch\ jetzt, jetzt ist das jetzt erst jetzt das nur jetzt ist und doch jetzt\ ist, nur jetzt und doch jetzt, jetzt das jetzt ist, nicht jetzt das jetzt

nicht jetzt ist ist jetzt das jetzt ist wenn es jetzt ist, nicht jetzt

wie es jetzt nicht ist, nicht jetzt wie es jetzt nicht jetzt ist, jetzt

das nicht jetzt ist ist nicht jetzt, jetzt nicht, jetzt noch nicht,\ doch jetzt das noch nicht jetzt ist wenn es jetzt ist, jetzt das jetzt\ nicht mehr jetzt ist wenn es jetzt ist und jetzt das jetzt ist wenn

es nicht mehr jetzt ist, dieses jetzt, erst dieses jetzt, nur dieses\ jetzt ist jetzt

– Max Bense, ‘Jetzt’, 1961 (Bense Ausgewählte Schriften Vol. 4 25).

The origin and profile of the study of the relations between natural and machine languages at the heart of the Large Language Models (LLMs) driving today’s artificial intelligence (AI) was not confined to the cognitive psychologists, logicians and computer scientists of post-war America, and therefore is not exhausted by either the original agenda of US AI science to ‘simulate’ human intelligence (as set out at the so-called Dartmouth conference, see McCarthy et. al. ‘A Proposal’), nor by the overdetermination of its research program and applications by military funding and, later in the century, by Big Tech investment. There were other, more heterodox explorations of machine intelligence, outside of the US and especially in creative, experimental applications, which did not display the same faith in the professional, scientific class to bring AI to life or to win an epistemic arms race. Instead, they embody an attempt to reconstruct intellectual production in the face of what was reality for much of the world in the aftermath of the war: a post-traumatic existence accompanied by a profound loss of faith in the practical application of Western epistemology. This explains the interest in the postwar period in West Germany, as we shall see, in tethering AI, information theory, and computing to aesthetic knowledge and aesthetic forms – literature, poetry, the visual arts, theatre – while noting in lieu of future discussion that in Continental philosophy aesthetics had long been the discursive site where the tensions between consensus and dissensus, reason and unreason, and logic and sensibility, were examined (Rancière 115–33). The exploration of machine intelligence in the postwar period was also a way to rethink and renegotiate a profoundly problematised relation to rationality and to knowledge itself. These, then, are my three intersecting foci for this piece: artificial intelligence, aesthetics, reparation.¹

To illuminate this intersection, I would like to take up and evaluate the work of the German cybernetician, information theorist, philosopher, design teacher, editor and poet, Max Bense (1910–1990), especially his theories aggregating around the idea of information aesthetics (informationsästhetik), and how these were formulated through the medium of literary experimentation. This essay will look at how the relation between natural and artificial languages played out across Bense’s own poetic artefacts, the literary communities he was involved in, and the avant-garde journals he presided over in the period after the war and into the 1970s. Unlike his contemporaries in North American AI, who viewed programmable machines as the simulation and automation of human intelligence, especially problem-solving and decision-making (McCorduck), Bense’s work and that of his circle – above all, the so-called Stuttgart School – saw them as instruments to create new forms of thinking and communicating through algorithmic patterning and structured randomness, and as a novel way to generate poetry and literature to appropriately engage with what they understood as the ‘artificial world’ (künstlichen Welt) (‘Kunst in künstlicher Welt’ 85). This vision was rooted in the post-war context, at ‘Stunde Null’, the zero hour: it was shaped by a desire to rebuild a world in ruins, to reconstitute more democratic and pluralistic ways of thinking from the fragments left behind by fascism. As such, Bense’s work offers a critical difference to the positivist impulse of early US AI, especially so-called ‘symbolic’ AI (Cardon et. al.). It also offers a counterpoint to the critique of technocratic culture conducted by Bense’s better known compatriots, most famously exemplified by Max Horkheimer and Theodor Adorno’s 1947 publication decrying the destiny of European reason, the Dialectic of Enlightenment. Max Horkheimer and Theodor Adorno saw that destiny embodied in a ‘thoroughly mathematized world’ in which thought is ‘reified as an autonomous, automatic process, aping the machine it has itself produced, so that it can finally be replaced by the machine’ (18–19). Instead, Bense’s work represents a theory of machine intelligence and machine aesthetics as a form of reparation to stitch back together the remains of the Enlightenment, including the fundamental role of mathematical thought in that tradition (Leopold), and, above all, to mend what he considered to be the broken connection between poetry and logic. What the early US AI researchers saw as the study of automata through determining the logical basis of all languages, natural and artificial, Bense and the Stuttgart School saw as the reunification of the estranged modes of intellectual production in ‘text theory’ (Texttheorie). Text theory and its implementation – what Bense would call ‘generative’ aesthetics, the ‘artificial production of probabilities of innovation or deviation from the norm’ (‘The Projects’ 57) – were for Bense methods to explore the question of intelligence across human and machines in a manner appropriate to the times. It may seem ironic to look at a then recently fascist state, one of the most violent regimes in history, for the glimmers of a democratic approach to technology, but this is indeed the task many intellectuals and artists at the time set themselves.

Bense’s intellectual trajectory encompassed physics, mathematics, philosophy, information theory, cybernetics, semiotics, literature and poetry. Reflecting this boundary-spanning between the human and technical sciences, his circle of collaborators and students shared the view that artificial and natural languages were complementary not opposed modes of study and creativity (indeed, ideas of co-reality, Mitwirklichkeit, and coexistence and collaboration, Mitwirkung, are common across Bense’s oeuvre). After providing some biographical details of this relatively unknown figure (at least in the Anglophone world²), I will follow the coordinates suggested above – intelligence, aesthetics, reparation – to set out the context and profile of his thought before focusing on some specific details of the literary experiments of his circle, including Bense’s own creative work. To conclude, I return to the post-war origins of AI to confirm its transnational and transdisciplinary nature. Finally, a note on the epigraphs to each section: they are examples of concrete poetry and literature from the Stuttgart School, and serve a number of functions: to illustrate the formal experimentation of the School and its principles for modern art, to echo some of the above themes I will be using to elucidate Bense’s vision of a new aesthetics, and to reflect the experimentation of the group in the essay itself. Some examples are intentionally left untranslated because, and true to the School’s intention, they should be read for form as much (if at all) for meaning – but I will translate or at least indicate where sense is in fact essential to interpretation.³

Stuttgart and Ulm

Trained in physics, Bense initially forewent a university career in the 1930s by refusing to comply with Nazi directives on his academic path and tried to emigrate to the United States in 1939 without success (Stewart, ‘The Higher Forms’ 41–42). Given his training, during the war he was sent to work at Hans Erich Hollman’s Electrical Engineering Institute in Berlin and then, after that site was bombed, to Georgenthal, Thuringia. After the war, it appears he was trusted enough by the Allies to be appointed mayor of that municipality. Finding himself on the wrong side of the iron curtain and under increasing scrutiny for his outspokenness, Bense managed to get to the West where he was to eventually join the Technische Hochschule Stuttgart in 1950 (hereafter TH Stuttgart). There he established a General Studies program (Studium Generale) and proceeded, among many other publications, to bring out the four volumes that would constitute his Aesthetica (c.1954–1960), each and all devoted in some way to producing an aesthetic discourse grounded in statistics, probability theory and information theory. As part of the Studium, a gallery was set up in 1957 on the same floor as the Institute for Philosophy (of which Bense was director) which exhibited concrete poetry, constructivist art, experiments in typography, etc., and became one of the first galleries to display and theoretically contextualise computer generated art.

Bense and his colleague (and partner) Elisabeth Walther founded and edited several influential periodicals, the magazine augenblick (1955–1960), the series rot (1960–1990) and the international journal of semiotics and aesthetics Semiosis (1976–1990). Bense also had a leading role in the long-running Grundlagenstudien aus Kybernetik und Geisteswissenschaft (Fundamental Studies in Cybernetics and the Humanities; hereafter GrStKG). Contributors especially to augenblick and rot were a who’s who of avant-garde literature of the time, including Gertrude Stein, Francis Ponge, Jean Genet, poets from Latin America Eugen Gomringer and Haroldo de Campos, and American Allen Ginsberg for good measure. Indeed, Bense valued his international connections, making numerous trips to the US from the late 1960s onwards, and maintaining important relations with writers, poets, artists and designers from Latin America (particularly Brazil), from Japan, and across Europe. The work collected and theorised in these journals included machine-generated art. Of rot, for example, Bense stated that it was not ‘limited to the natural production methods of a creative individual but also includes the artificial productions of electronic computer systems’ (qtd. in Robillard 10). The nineteenth edition of rot in which Bense provided a short explanation of the project of generative aesthetics was devoted to computer graphics.

The Stuttgart School itself (sometimes Stuttgart ‘circle’, Kreis) was an experimental literary group of the 1950s and 60s. As part of this experimentation, its members produced computer-generated texts using stochastic modelling based on the American electrical engineer and mathematician Claude Shannon’s information theory (1948) which, in turn, was inspired by Russian mathematician Andrei Markov’s 1913 statistical analysis of Aleksandr Pushkin’s 1833 novel in verse, Yevgeny Onegin (more on this below). Using a Zuse Z 22 mainframe computer, then newly available at TH Stuttgart, simple algorithms were written to generate unexpected textual associations. Stuttgart mathematician Theo Lutz, for instance, reorganised Franz Kafka’s 1926 novel Das Schloß (The Castle) at Bense’s suggestion using a basic algorithm randomising sections of the novel to form new sentences and sentence chains, an example we will return to. In a review article of 1960 published in GrStKG, Bense, furthermore, saw the development of information aesthetics (and therefore text theory) as much a result of such experimentation in generative textual practices as an application of information and communications theory to the writing of texts. Experimental literature was the combination of ‘reflections on structural and Gestalt theory, communication research, perception theory, general semantics, and occasionally even statistical linguistics’, with ‘visual, material, serial, stochastic, automatic, and vibratory techniques’ (‘Movens. Experimentelle Literatur’ 123).

Bense and Walther also taught at the post-war, neo-Bauhaus design school, the Hochschule für Gestaltung, Ulm (hereafter HfG). The HfG was created with the support of Marshall Plan funding by the Scholl Foundation. That Foundation served the memory of the White Rose resistance group murdered by the Nazis in 1943 (see Scholl). The first version of the HfG was an adult education centre (Die Ulmer Volkshochschule). Well-known political and literary figures played a key role in devising the original curricula of the centre, including Hans-Werner Richter and others associated with writer’s collective, Gruppe 47 (Frey; Braun). Bense gave his first lecture at the centre in 1952 (Oswald et. al. 11). A year later when the centre became a design institute under the direction of concrete artist, architect, and former Bauhäusler, Max Bill, the Information Department remained as an element of the original concept. Bense and Walther were critical figures in that small but energetic group (see Bonsiepe, ‘A Personal Retrospective’). Staff and students experimented with journalism, documentary practice, and radio and film, to analyse the circulation of messages that constituted public communication and opinion from an information theory perspective. In a fascinating parallel to Jacques Derrida’s (in)famous statement, ‘there is nothing outside of the text (il n’y a pas de hors-texte)’ (158), Bense once declared as early as 1960 that, ‘[o]utside a text there is no flow of language, no flow of thought, no flow of information’ (Außerhalb eines Textes gibt es keinen Sprachfluß, keinen Gedankenfluß, keinen Informationsfluß; Bense Aesthetica Vol. IV 79). The argument was that to restore trust in the public sphere a thoroughgoing scientific understanding of communication was needed. To that end, recent developments in statistics, control theory (cybernetics), and information theory were adopted by the HfG. The department, however, and ironically given its mission, never had more than a small number of students and was eventually absorbed into Alexander Kruge and Edgar Reitz’s Film Department in the early 1960s.

A severe critic of any moment the country attempted to ignore its recent past, Bense’s political outspokenness would be something of a trait across his career in the West leading to several very public ‘affairs’ (Stewart, ‘The Higher Forms’ 42; Jörg-Rössler; Bonitz). But while he would not countenance remnants of fascist ideology and ever highlighted the idea of ‘Engagement’ throughout his writing (for example, Artistik und Engagement), he was also keen to dispute those who linked Western political catastrophe to the Enlightenment tradition, and, for that matter, Bense never ascribed to what he called the Jacobin tradition of democracy; ‘one acts socially, but decides individually’ he once put it (‘Engagement und Experiment’). In holding this view, Bense’s thinking offers a counterpoint to his better-known fellow nationals and returned expatriates, namely Martin Heidegger, Jürgen Habermas, and Adorno, and provides a more complete picture of this period of German thought.

Concrete poetry / *Konkrete Poesie*

ismus ta ti istisch ti ta ti ta to ti ti to istisch

ta ta to ta istisch ti ismus ti ti to ti istisch

ta ta ismus to to ti ismus ti to istisch ta ti

ta istisch ismus to to ta to ti ta to ta ta

to ismus ta ismus ti to to ta ta to istisch ti

ta to ismus ta ismus istisch ti istisch ta ta to

istisch ti ta ti ti ta to ti ti ti to ta ismus to

to istisch ti ti to ti to ti ta ta ti to ta ti

ti istisch ti ismus ti ta to istisch to ta istisch

to to to ti ta to to ta ti istisch ta to ismus

ti ismus to istisch ismus ta istisch ti ti to to

ismus ta ti ti ta istisch ti istisch ta ismus ta

to ta to ta to to ta to ismus to to ismus

to ta ti ta istisch ti ta ta ti ta istisch ismus

to ta istisch to to ta ti to to ismus ti to ti

ismus ti ti ti ismus ta ismus ti ta istisch

– helmut heissenbüttel, deutschland (rot, no. 21, 1965).

rio

roi

orior

– max bense, deutschland (rot, no. 21, 1965).⁴

As noted, Bense was a key figure in the concrete poetry movement, which had roots in Latin America as much as continental Europe. The Swiss-Bolivian poet Eugen Gomringer did much to define the movement, and Haroldo de Campos (lifelong friend of Bense) and his Noigandres group from Brazil made significant contributions (Drucker). Above all, the concrete poets were interested in the material, structural components of language. This was embodied by interest in the shape or form of the poem, its typographical layout, and its ‘internal’ laws; how, for example, a word or sequence of words appeared next to others in terms of graphic and phonic assonance, and how ‘found’ texts could be broken down and recombined, rather than words being windows, so to speak, to the world. The experience of a poem was therefore to be at once elemental and multifaceted. This went for subject matter too: the concrete poets did not differentiate between fine and applied art, high or low culture, human or machine languages, and often appropriated the graphic idioms of commercial industry, of advertising and signage, as well as pulling text from the media (for example, snippets of radio and print) and from technical manuals and scientific reports. In the twenty-first edition of rot, collecting concrete poetry from around the world, Bense summarised the intentions of its program. Concrete poetry, he wrote,

… is a form of poetry that does not generate the semantic or aesthetic meaning of its elements, such as words, through the usual formation of linear and grammatically ordered contexts but rather reflects on visual and spatial (flächige: or ‘planar’) connections. it is not the sequence of words in consciousness that is the original constructive principle of this type of poetry, but their coexistence (miteinander) in perception (‘konkrete poesie’, 44).

Explicit in this turn to the concrete and to perceptual coexistence – to the structures and variety of linguistic materials – was a critique of a particular view of authorship. This was a direct challenge to what Bense and his Stuttgart colleague Reinhard Döhl called in their 1964 manifesto of experimental/concrete poetry, Zur Lage, the ‘mystic and metaphysical windbags (Schwadroneurs)’ of previous poetic traditions (Zur Lage). They had in mind here the extremities of the Romantic tradition, and no doubt the appropriation of that tradition in the aesthetic preferences of the Third Reich, especially so-called ‘Steel Romanticism’ (see Dennis 176–97). That kind of subjective, expressive style of authorship had, according to Bense and Döhl, been a vehicle for ‘questionable ethical content’ and ‘ideological nonsense’. It was now to be replaced by the ‘atheist’, the ‘rational and methodical author’, who instead explored the ‘microaesthetic structures’ (mikroästhetische Strukturen) of language. While not drifting too far from the longstanding Germanic privilege given to poetry amongst the arts, this was nonetheless a call to grant autonomy to language in all its variety and detail, and in all its manifestations. The purpose behind the study of the independence of communicative phenomena was to counter the ideologically saturated language of the past. It was the contemporary stochastic and informational sciences which were the key to this repair work because it was these sciences that could be deployed to reconfigure aesthetics so that it could pertain to all forms of communication (as information aesthetics). In this spirit, Bense and Döhl went on to summarise six tendencies of experimental art:

1. Letters = type arrangements = letter images

2. Characters [or signs: Zeichen] = graphic arrangement = type-images

3. Serial and permutational realisation = metrical and acoustic poetry

4. Sound (Klang) = tonal (klangliches) arrangement = phonetic poetry

5. Stochastic and topological poetry

6. Cybernetic and material poetry.

These tendencies were enumerated to describe experimental literature but were equally relevant to any linguistic material. Instead of envisaging what Bense called the ‘exact’ sciences as a source of the dehumanising characteristics of German ‘ideology’, akin to the comment we have quoted above from Horkheimer and Adorno, the Stuttgart School saw instead a path to reconstructing aesthetics, rationality, and language. Explaining the phenomenon for the readers of Deutsche Zeitung in 1961, the Stuttgart School poet and novelist Helmut Heissenbüttel, who made ‘found’ texts and quotations an essential part of his literary practice, replacing sequence and causality with associations and correspondences, described the goal of concrete poetry was to reduce textual practice to

… words that are as uninflected as possible, to word sequences or simple formula chains (Formelketten), insofar as they represent simple word relationships such as addition, negation, inversion, or a question. There is a certain analogy to the basic formulae of symbolic logic. Everything that is indeterminate, complicated, or obfuscated must be disregarded. (‘Konkrete Poesie’)

A decade later, in an issue devoted to the concrete poetry movement in the literary journal, Text + Kritik, Heissenbüttel adds to the idea of reduction that of exceeding ‘media limitations’ (medialen Begrenzungen) (for another contemporary explanation of the movement, see Kopfermann). After reduction to ‘logic’, passing words and word sequences through different media – the visual, the aural – allows for the avoidance of the ‘foundational Western schema of subject-object-predicate’ in favour of ‘opening and grinding (öffnenden und verschleifenden) new syntagmatic experiments’:

Typographic and acoustic border crossings (or perhaps more correctly: eccentricifications [Exzentrifizierungen]) now also serve as a means of establishing new or at least unusual syntactic bonds: for example in the chaining of words by changing letters, in the use of rows, permutations, chains of variations, etc.; but also in the transition to echoing sounds, in the use of speech acoustics as a kind of musical composition building block. (Kopfermann 19)

Heissenbüttel concludes his 1971 notes, which also formed an introduction to a selection of his own poems (a fragment reproduced below), with a claim common across Stuttgart School experimentation: ‘moving away from conceptual rationality does not mean abandoning rationality altogether. On the contrary. But we must first try out where rationality can be applied anew’. (21) For Heissenbüttel concrete poetry is part of a new way to orientate oneself linguistically in the world. Heissenbüttel’s phrase I have translated as conceptual rationality is begrifflicher Rationalität. In German, concept (Begriff) derives from the verb greifen, to ‘grasp’, and holds a significant place in German thought and is especially associated with Hegelianism. Here, though, Heissenbüttel is using it as a counterpoint to concrete poetry’s own rationality which is not interested in grasping the essence of something, or in Hegel’s terms, the philosophical attempt to search behind or amongst representation (Vorstellung) for the concept (Elements of the Philosophy of Right 27). Heissenbüttel is therefore negotiating a fine path between the legacies of idealism and, on behalf of the Stuttgart group, the focus on all the dimensions of language available to the poet. As also argued by Bense and Döhl, breaking down and recombining language – finding language everywhere, so to speak – is part of the work of reconfiguring rationality.

Intelligence

77

98

777\ 7

777 777

98 . 98 .

7

77 . 9 . 8 .

9 898

989 898

898

90

9

______________

7 . 7 . 7 . 7 . 7 .

8

9.

– Helmut Heissenbüttel, no title (excerpt) (Text + Kritik 1971, 17).

The term ‘artificial intelligence’ was coined in 1955 in a submission to the Rockefeller Foundation to fund a summer research series at the Ivy League institution, Dartmouth College. AI was partly so named to exclude the founder of cybernetics, Norbert Wiener, from coopting the field for his eponymous new science (McCarthy, ‘Review’; Kline). The authors of the submission are now household names, in computing and cognitive science at least: John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon. The research vision articulated in the submission and explored the following year at the summer seminar was essentially ‘symbolic’ AI, that is, a mimetic approach to intelligence in automata: ‘The study of Artificial Intelligence is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it’ (McCarthy et al. 2). For the Dartmouth group, intelligence meant using language, forming general concepts, problem-solving, decision-making, the ability to learn, and, indeed, creative thinking (through randomisation).

The approach Bense and the Stuttgart School took to artificial intelligence of course echoed the developments in the US. Indeed, along with contributions from Augusto and de Campos, Ronaldo Azeredo, Gerhard Rühm, Heissenbüttel, Gomringer, Ponge, and Bense himself, the first exhibition of concrete poetry in Germany held at Bense’s TH Stuttgart gallery in 1959 included text by Claude Shannon (Walther 353). Bense was an early reader of Wiener’s Cybernetics, Or Control and Communication in the Animal and Machine, which was sent to him in 1949 from the US by the newly relocated Hollmann, the head of the lab Bense worked for during the war. In the mid-50s Bense invited Wiener to speak at both Stuttgart and Ulm, having previously written an essay on cybernetics for the journal Merkur (‘Cybernetics, or the Metatechnology’). Aside from the writings of Shannon (and Warren Weaver), Bense closely followed those working in cognitive science and communication engineering, such as Donald M. MacKay and Colin Cherry.

But Bense and the Stuttgart School did not pursue the intersection of art, information theory, and computer programming with the goal of disarticulating human intelligence and reconstituting it in languages for machines. For Bense, information theory as the basis of programmable text experiments was part of redefining communication for what he called ‘technical civilisation’ or ‘technical existence’. This is the way Bense described this existence in 1949:

The world that we inhabit is a technical world. It is the world of processes, functions, flight paths and station stops, the world of machines and calculations, of gears, noises, factories, and transmissions, the world of technicians, engineers, physicists, experts, specialists, professors, secretaries, and institute directors, the hardly fathomable world of unions, guilds, firms, laboratories, industries, canals, cities, mine shafts, depths, and heights, the world of timetables for trains and electrons, the world of the masses forever knocking at the gate – and the world of a calm intelligentsia (Intelligenz) encased by the thin but unyielding walls of respon­sibility for everything that belongs to this world, protected from any confusion that might disturb the creative process or infect the mind (Geist) with distrust. This world is no mere potentiality, and it is not a draft that can be revised and rejected, a sketch on a piece of paper. It is an undeniable reality; it is reality outright. In our time, being imposes itself on our interior and exterior existence neither in the shape of nature nor in that of culture. We inhabit not land­scapes and gardens, not houses on sloping hills or in bright glades. We inhabit a network of visible and invisible functions and rela­tions, structures and aggregates made of metal and artificial stone that have taken on names like towns, cities, countries, and con­tinents. Technics (Technik) concerns us. We both love and hate its forms. (‘Technical Existence’ 49)

This remarkable and very specific emphasis on all the bits and pieces of contemporary reality, people, things, environment, is what Bense meant by ‘concrete’. Bense in fact took the word from Hegel, who made a distinction between the concrete and the abstract (‘konkrete poesie’). Abstract thinking for both Hegel and Bense lifts a few key features from the complexity of reality to define its object, in the process rendering the other features forgotten, or mute; while concrete thinking, their mutual preference, attempted to do justice to all aspects of the phenomenon, to multiple perspectives. In ‘Who Thinks Abstractly?’ c.1808, probably written for a Jena newspaper but unpublished (Gray), Hegel argued those who think concretely rather than abstractly are the ‘educated’. The locution Hegel used is die Gebildeten, that is, the ability to see all sides of a phenomenon, from every view (Bild) and angle. This idea of endeavouring to account for everything, not in the sense of control but in the sense of a fundamental regard, even empathy, is critical to Bense’s view of the artificial world (künstlicher Welt). It led to what he also called in the 1951 Merkur article via an extrapolation of cybernetics, ‘metatechnology’ (‘Cybernetics, or the Metatechnology’), a ‘paradigm’ to conceive the tessellation of technology across all domains of being.

Unlike some of his contemporaries, most notoriously Heidegger, there was no desire on the part of Bense to reverse or escape from the technological state of the world; no suggestion that, in our supposed condition of being dominated by technics, we needed saving (as per the infamous quote from the Der Spiegel interview with Heidegger in 1966: ‘Nur noch ein Gott kann uns retten’). The artificial world for Bense was the contemporary lifeworld. In terms of facing and making sense of that world, there was no attempt on the part of Bense or the Stuttgart School to return to the subjective, idealistic poetics of the Romantic tradition, or to that tradition’s intense appreciation of nature as expressed in Sturm und Drang. The inner-focused, expressive domain of aesthetic experience was recoded as a trans-subjective experience underwritten, quite literally, by information theory: information is what is common to natural and machine languages, to the design and composition of the artefacts of high and low culture, to the messages and signals of both the public and private spheres, and to the argots of the spiritual and mundane. And while the results of Bense and the School’s experiments were often as difficult to read and interpret as any avant-garde project, the automated language trials of Stuttgart were intended to explore public communication or, more concretely, democratic communication. This, it could be imagined, was a difficult and problematic task in Germany post-1945 and its landscape of trauma, its complex politics of guilt and atonement, of exile and return, but also of denial and forgetting.

There are three key terms captured in this rich description of the artificial world opening ‘Technical Existence’ that are further explored by Bense throughout the essay and already highlighted in German in the text – Intelligenz, Geist, and Technik. Bense redefines them individually and in relation to each other. Intelligenz is intelligence, naturally, and will become the word used to make the phrase artificial intelligence, künstliche Intelligenz. But it also can mean intellect (and collectively, the intelligentsia), a meaning which, in turn, echoes that most loaded of words in the German language, Geist. Geist is ‘spirit’, but also ‘mind’, ‘genius’ and ‘ghost’. It can have epochal associations, as in the phrase describing the spirit of the times, Zeitgeist. It connotes the higher orders of human activity, and is used, for example, in the German version of the ‘human sciences’, Geisteswissenschaft. Certainly, in historic use the word evokes a counterpoint to die Technik (‘technics’, connoting both technology and technique). This is the new relation between them Bense establishes: far from being displaced from the spirit and intellect, technics is now the very object of the work of the mind (Geist), and this work is conducted by a redefined idea of ‘intelligence’ (Intelligenz). Intelligence is not abstract thinking (or ‘instrumental’ reason for that matter), nor a phenomenon that lies in a brain or a mind which can be analysed and then replicated in machines. Rather, it is a scientific attitude that attempts to address the technical world through the medium of language. Intelligence is a sensitivity to what he called ‘linguistic action’. These ideas inform Bense’s characterisation of the ‘engagement’ of the experimental writer, which he defined as ‘existing through writing’ and as commitment to a ‘humane use of writing’ (einen humanen Gebrauch vom Schreiben) (‘Engagement und Experiment’).

Even in 1950 this way of thinking informs Bense’s review of Horkheimer and Adorno’s Dialectic of Enlightenment (and Adorno’s Philosophy of New Music). After summarising Horkheimer and Adorno’s argument – in attempting to extinguish fear and superstition and establish the ‘rule of man’, the Enlightenment and the narrative of progress that underwrites it instead relapsed into mythology, nationalism, and paganism – Bense went on to argue this vision was a one-dimensional interpretation, repeating the old trope of judging the modern world solely in terms of intellectual history (geistesgeschichtlich). According to Bense, this trope, a ‘typical Bourgeois concept’ he says, inevitably deprecates mathematical, scientific, and technical enterprise. The Hegelians in California missed that the Enlightenment was an empiricist movement as much as one of rationalisation and domination of nature – a movement Bense argues that was geared to (and still may lead to) a ‘democratisation’ of ‘results’ (‘Demokratisierung’ der ‘Ergebnisse’) (‘Hegel und die Kalifornische Emigration’ 122). The point here is that Bense is attempting to restore Enlightenment reason and values by appealing to the multiplicity of conceptual and practical vectors, a ‘concrete’ view of the tradition in other words.

Aesthetics

‘In der Kunst sind wir stets im Feld des “Gemachten”, nie des “Gegebenen’’’. (‘In art, we are in the field of the “made”, not the “given’’’.)

– Bense (1956, 35).

NICHT JEDER BLICK IST NAH. KEIN DORF IST SPAET.\ EIN SCHLOSS IST FREI UND JEDER BAUER IST FERN.\ JEDER FREMDE IST FERN. EIN TAG IST SPAET.\ JEDES HAUS IST DUNKEL. EIN AUGE IST TIEF.\ NICHT JEDES SCHLOSS IST ALT. JEDER TAG IST ALT.\ NICHT JEDER GAST IST WUETEND. EINE KIRCHE IST SCHMAL.\ KEIN HAUS IST OFFEN UND NICHT JEDE KIRCHE IST STILL.\ NICHT JEDES AUGE IST WUETEND. KEIN BLICK IST NEU.\ JEDER WEG IST NAH. NICHT JEDES SCHLOSS IST LEISE.\ KEIN TISCH IST SCHMAL UND JEDER TURM IST NEU.\ JEDER BAUER IST FREI. JEDER BAUER IST NAH.\ KEIN WEG IST GUT ODER NICHT. JEDER GRAF IST OFFEN.\ NICHT JEDER TAG IST GROSS. JEDES HAUS IST STILL.\ EIN WEG IST GUT. NICHT JEDER GRAF IST DUNKEL.\ JEDER FREMDE IST FREI. JEDES DORF IST NEU.\ JEDES SCHLOSS IST FREI. NICHT JEDER BAUER IST GROSS.\ NICHT JEDER TURM IST GROSS ODER NICHT JEDER BLICK IST FREI.\

– Theo Lutz, ‘Stochastic Text’, Kafka’s The Castle (excerpt) (1959).

At least since Kant’s Critique of Judgement (1790), in German philosophy aesthetics has served as a field for negotiating the tensions between pure and practical reason, subjectivity and objectivity, order and disorder, intellect and sensation, and between consensus and dissensus. Kant thought he could espy in the aesthetic domain the general conditions of agreement; if we can find the principles for agreeing on something seemingly so subjective as aesthetic phenomenon (on something so subjective as ‘taste’) then this may provide an analogy to the moral world to guide the practical applications of reason. For others, aesthetics was to build no such bridge: aesthetic autonomy was itself a model for both political and epistemic autonomy based in radical dissensus: the right, so to speak, to create different worlds (for a crisp articulation of this position see Horkheimer Critical Theory 273–90; and for an extensive one, see the entirety of Adorno’s Aesthetic Theory).

Bense waded into this long and complex tradition (for background see Holt). His central concept of ‘information aesthetics’ attempted to bring together stochastic processes, the essential propositions of information theory, and aesthetic experience. Drawing on Shannon’s information theory, Bense proposed that aesthetic phenomena – texts, artworks, poetic forms – could be analysed in the information theoretical terms of entropy, redundancy, and statistical deviation. But unlike the goal of communication engineering to reduce noise in a signal for a more efficient and legible transmission, his project sought to reveal, rather than eliminate, the aesthetic potential of artificial processes. This culminated in his theory of ‘generative aesthetics’, defined as the ‘artificial production of probabilities differing from the norm using theorems and programs’ (‘The Projects of Generative Aesthetics’ 57). For Bense, this generative capacity was not merely technical; it was philosophical and even ethical. It offered a way to create meaning within an artificial world (künstlichen Welt) – a world not simply made of machines but constructed after the collapse of epistemic and moral certainty. Bense also overlaid semiotics – especially as derived from the American pragmatists, Charles Sanders Peirce and Charles W. Morris (Holt) – with information theory, but here for the sake of brevity I will focus only on information and its relation to language.

The origin of the mathematical basis of both the Stuttgart experiments in artificial texts and the LLMs which power contemporary generative artificial intelligence lies in the statistical tables of Russian mathematician, Andrei Markov. The original tables were generated from analysing Pushkin’s novel in verse, Yevgeny Onegin. Markov took a 20,000-word extract from the poem and laid out the arrangement – as they occurred, exactly – of vowels and consonants (omitting punctuation and pronunciation modifiers) and weighted the relative frequencies between them to give a probability of likely sequential occurrence. Put simply, if we did this in English, one would find T-E is much more likely to occur than T-T. This can be done with strings of syllables, words, clauses, and so on; notated as bigrams, trigrams, etc. Today, neural nets can work in dimensions where the n of n-gram is 25,000 plus, and along with attention mechanisms (algorithms which guide relevance) allows for the generation of highly plausible outputs. Markov presented his results at the Royal Academy of Sciences in St. Petersburg in 1913, and as published at least, it was a dry lecture, concluding not with any grand declaration but with a number (Markov; Link). Of course, Markov was not interested in establishing insights into Pushkin but in creating new parameters for probability theory. Indeed, the lecture marked the moment in which probability theory – and therefore, ultimately, statistics – shifted from the basic idea of equal chance under ideal conditions (a perfectly neutral coin toss, for example) to the idea of connected events. Quite simply, in an organised phenomenon like language there is no equiprobability of one symbol or group of symbols following another; there are structures, patterns, and expectations that have an influence of what occurs next. In the tradition of Markov’s analysis of Pushkin, in the hands of the early computer scientists who drew upon probability theory and statistics, natural language became the site through which the modelling of machine language was conducted: the artificial, i.e., the stochastic, pattern-like nature of language could tell us something about human language, and human language could assist us in creating stochastic models.

As already suggested, this modelling was most famously and influentially undertaken by Shannon to define ‘information’. In 1948, Shannon published his work on the statistical theory of communication which set out the destiny for the meaning and application of information theory for the rest of the twentieth and twenty-first centuries. Working for Bell Labs, Shannon’s interest was in signals, specifically how any chosen message can be encoded and transferred down a channel to, optimally, reach its destination relatively intact. He was not, therefore, interested in what the message meant (semantics) but in that process of selection itself: a symbol must be selected from a set of possible symbols. As information is defined relative to the probability of decrease in uncertainty, rules and patterns determining selection help limit uncertainty (this is the case in all episodes of signal transmission, schematically: sender, message, receiver).

To grasp the stochastic nature of any message, Shannon, like Markov, analysed natural language artefacts, selecting letters at random from pages of a book. He started by choosing letters where all symbols are ‘independent and equiprobable’ (such is ‘zero-order approximation’):

XFOML RXKHRJFFJUJ ZLPWCFWKCYJ

FFJEYVKCQSGHYD QPAAMKBZAACIBZLHJQD

But then he progressively moved through digrams and trigrams to word units, getting closer to what an English statement may look like; Shannon finally was able to generate recognisable English text:

THE HEAD AND IN FRONTAL ATTACK ON AN ENGLISH WRITER THAT THE CHARACTER OF THIS POINT IS THEREFORE ANOTHER METHOD FOR THE LETTERS THAT THE TIME OF WHO EVER TOLD THE PROBLEM FOR AN UNEXPECTED.

The statistical approach to messages is, therefore, both analytical and generative. It can be used to predict the probabilities of words and sentences (but also the composition of the visual field) in natural and synthetic languages, but also to engender both likely and unlikely associations.

Bense and the Stuttgart group performed similar stochastic experiments in automating language but in an obviously more avant-garde sense: not by narrowing the gap of recognisability but by exploring the widening of that gap, an exploration, then, of randomisation and its limits. An example, already mentioned, is Theo Lutz’s ‘random texts’. Lutz’s texts followed the basic idea of formulating some grammatical rules (the algorithm) but then selecting the content in an arbitrary fashion from a given inventory. Lutz chose 16 nouns and 16 adjectives from Kafka’s novel The Castle and added a selection of conjunctions and pronouns. He then encoded this language material on punched cards. The algorithm programmed for the Zuse Z 22 then combined them into a large but relatively simple series of grammatically correct main sentences. The result was text of about 70 lines printed out on a teleprinter (sample text in section epigraph, above).

For Bense, Markov chains and Shannon’s information theory allowed him to evolve a theory of general communication but also to make differentiations between, for example, what he termed classical and non-classical texts. By the former he meant the literature in which the rules (of genre, for example) are relatively well defined and understood, and in which, therefore, the set of possible message selections are also relatively determined (a recurring example of his, as with other members of the Stuttgart School, is the grammatical ‘rule’ of subject-object-predicate). By the latter Bense effectively means modernist literature; his favoured exemplar is, above all, James Joyce, but he also consistently references Samuel Beckett, Ponge, Gomringer, and of course the German experimentalists of his own circle, inter alia. In the hands of the modernists, text is a lot less predictable; or, put differently, the ‘logic’ of the modernist text is created through syntactical, material experimentation with language. Stochastic processes and information, therefore, serve a dual function, analysis and synthesis: ‘The fact that general text theory is built up from the components of informational aesthetics and statistical text analysis not only opens up the possibility of breaking new ground in the analytical recognisability of texts, but also in the synthetic constructability of texts’ (Aesthetica Vol. IV 113).

What joins the classical and the non-classical, so too the logical and the poetic, is Bense’s take on Shannon’s notion that messages are selected from a bank of possible message choices. In the case of Stuttgart’s understanding of linguistic material, the avant-garde is charged, so to speak, with expanding this set. In his 1960 book, Programmierung des Schönen (Programming the Beautiful) (where, not incidentally, to the idea of the technical or artificial world he adds the world as ‘process’), Bense conceives of this set as the ‘reservoir’ of language in terms of information theory but also in terms of the cybernetic concept of equilibrium:

Language will always have to be imagined as a reservoir of signs (ein Reservoir von Zeichen) and their complexes, and every text that is formed from them appears in it as a determining principle, as a statistical super-sign (als statistisches Superzeichen) that is to be interpreted as a cell of a new structuring or classification, but also as the starting point of a disturbance of the equilibrium… [of] the reservoir. It is a characteristic of great poetry and literature that for a long time it inhibits language for other great poetry and literature in certain respects, cutting off creative possibilities. One must then resort to imitation (Nachahmung), to a fixed style, or shift the work of art from poetry and literature back to the realm of pure text, where the reservoir, language (die Sprache), can be enriched again, for only enriched language can lead to new aesthetic information and meanings, and what appears to be experimental creativity (Artistik) is in truth creation on a deeper level… (Aesthetica Vol. IV 85–86).

Such is how Bense both aggregates and reconciles all the different modes of literary practice, prose and poetry, classical and non-classical, the logical and the random. Following the inspiration of the American pragmatists (especially Charles Sanders Pierce’s semiotics) and Shannon’s information theory, the conclusion Bense reaches is that the appropriate mode of present-day knowledge is not culturally specific language but a general semantics. As such, Bense differentiates between classical aesthetics and the contemporary aesthetics of technical civilisation by defining the former as concerned with objects in themselves while the latter is concerned with what is communicated. ‘In classical aesthetics’, he wrote, echoing Hegel’s phraseology, there is ‘something in and of itself that is beautiful (schön): the moon, sun, wind, a rose, scent, a feeling, etc. In modern aesthetics things only become beautiful through the sign that one finds for them: through the sound, the verse, the image, the metaphor, through arrangements, rhythms, metrics, perspectives’ (Aesthetica Vol. II 30).

Thus, Bense’s project was to clarify the nature and role of post-classical aesthetic knowledge which he also called ‘transclassical aesthetics’. In the context of both Stuttgart and Ulm, this meant aesthetic production and experience no longer belonged exclusively to the artist and to the art object, but was a broader phenomenon, as we have noted, encapsulating both human and machine, and all ‘communicative’ artefacts (art, design, poetry, advertising, political messaging, and radio and film, etc.). More concretely, post- or trans-classical aesthetics is based on a distinction made by fellow German philosopher and cybernetician, Gotthard Günther, who spent much of his career in the US. Günther differentiated between classical and transclassical machines, the former belonged to the world of work and energy, while the latter belonging to the new world of electronics, and therefore of communication and information (Ungehorsam der Ideen 31–35) The idea was that, as media of consciousness, both art and engagement (and recalling here the intentions of the Information Department at the HfG) were grounded in language:

Creativity and commitment (Artistik und Engagement), understood as intellectual (geistige) actions, have their medium in consciousness and thus in language. Creativity and commitment operate essentially linguistically (sprachlich) (Die Realität der Literatur 17).

To further define aesthetic information Bense leant on Shannon but also the American mathematician George David Birkhoff’s measure of relative order and complexity, expressed in the formula M = O/C, where M is ‘aesthetic measure’, O is order, and C complexity (see Birkhoff 6). Put simply, redundancy is order (or pattern). Less complexity therefore means more predictability (pattern); more information, less certainty. At the HfG, these ideas were applied to designing, in Stuttgart these principles were applied to literature. At Ulm, exercises were held in investigated the points at which objects and images gained or lost their shape or form (Gestalt) in terms of informational value: what ‘amount’ of information, students were asked to investigate, was needed to perceive and/or use objects? In the design artefact the communicative function is obviously more pronounced, therefore pattern or predictability is directly related to functionality and purpose; while at the other end of the spectrum lies the informationally maximal gestures of art, particularly avant-garde art, that is, they are more difficult to ‘work out’. The latter was crucial for Bense because for him the new rationality, ‘creative’ rationality (Ungehorsam der Ideen 95), must embrace not exclude the unpredictable, the random, or the oblique. The other idea Bense took from information theory and cybernetics was the relation of entropy to information, siding, in this case, more with Norbert Wiener than Shannon, insofar as Bense understood physical states as entropic and aesthetic states as fundamentally negentropic (Aesthetica Vol. IV 19–20; Walther 335). For Bense, it can be proved, so to speak, that aesthetics counters the entropy of the material world.

Emerging from these studies and applications of information theory, Bense’s most distinctive contribution to aesthetic theory was his concept of ‘generative aesthetics’ (and all the other ideas which are versions of it: text theory, transclassical aesthetics, stochastic texts, artificial poetry, etc.), an approach that views artistic creation as a structured process governed by rules and algorithms in contrast to the conventional notions of art as a spontaneous, expressive activity, and is defined by the attempt to persistently hold together rules and variation, order and randomness.

Reparation

no

nai

non

nain

none

naina

nonei

nainai

nonein

nainain

noneinai

nainainai

noneinain

nainainain

ou uo

no on

on no

uo ou

– Franz Mon, ‘Akklamation der verneinung’ (Text + Kritik 1972, 23).

Bense’s own experimental practice was extensive. The poem that is the epigraph of our introduction, ‘Jetzt’ (‘Now’), was first published in rot no. 6, and was a reduction, so to speak, of a chapter from Hegel’s Phenomenology of Spirit. According to Walther, Bense occasionally referred to this work of his as ‘word jazz’, which makes sense in that one of its published forms (not reproduced here) was a dynamic, spiralling shape (Ausgewählte Schriften Vol. 4 69). The concrete poem ‘Teile’ (‘Parts’) was published together with Paul Wunderlich’s lithographs ‘20 Juli 1944’ in the fourth edition of rot, 1961. The hashed together text was simultaneously concrete and political, referring to the attempt on Hitler’s life (Operation Valkyrie) and the subsequent death penalties given to the conspirators involved. As a note to the text, Walther includes a leaflet written by Bense towards the end of the war which had been kept by fellow physicist and philosopher (and forerunner of machine art, who become a well-known science fiction writer) Herbert W. Franke, and, as Walther comments, forms the basis of some of ‘Teile’. The leaflet very clearly sets out his political position against the regime and highlights his Enlightenment cosmopolitism (see appendix).

Perhaps the most extraordinary of all Bense’s creative works, in my mind at least, was his radio play, belonging to the genre known as Hörspiel in Germany (Mandel 145–63). An initial text of the play appeared in rot no. 11, 1963, entitled ‘vielleicht zunächst wirklich nur. monolog der terry jo im mercey [sic] hospital’. It was supplemented by writer, radio playwright, and translator Ludwig Harig, who had extensive contact with the Stuttgart School and used found material and collage techniques in his own radio plays. Bense also devoted a short essay on Harig’s Hörspiele (Die Realität der Literatur 125–28). Der monolog der Terry Jo was fully produced and broadcast in 1968 by Saarländischer Rundfunk on 11 September 1968 (Bense and Harig 532). Bense himself outlines how the piece came about and the automatic language techniques used in an essay of 1971 on computer texts (Die Realität der Literatur 74–96), stating the experimental play summarised the project of the institute at Stuttgart insofar as it combined ‘machine, automatic, intuitive, and human language generation…’ (92).

It was based on a true story from 1961, in which an American family, the Duperraults, hired a boat to tour the Bahamas; the captain turned on them and murdered all but the youngest member, Terry Jo, who managed to take to sea on a life float and was eventually picked up. In hospital, she slowly reconstructs the trauma, fragmented and painful as her memories are (hear the Radio Bremen production: Bense and Harig 1988). Her story is reconstructed through interlocutors (investigators and hospital staff) and through contemporary news stories. The computer-generated text was carried out by the mathematician and student of Bense, Siegfried Maser, on a repertoire gleaned from media reports on the violent tragedy, and proceeded in the same approximative logic of Shannon, slowly gaining the statistical ‘sense’ found in natural language. At least initially, these approximations, from vowel and consonant distributions through words to phrases, were given a voice (that of Terry Jo) through a vocoder (for the ideas behind the original vocoder, see Dudley), creating a weird, ethereal set of sounds. There are six such transforms and I quote the first three:

Fyuiömge-sevvrhykfds-züeä-sewdhf-mciöwzäikmbw-aycfjtcuä-hwlgtüamöozqIspbrgeca-vdeüihyiwr-dxe

h-rahhueber--sh-dfnupz-cun-ikirae-rn-tws-fonnrtücn--dz-nedre-hoIrikma-ruekeomah-ü-t-Ishe-seab-ur-wh-ef

has-wirkleb-stion-und-füsse-etwas-dochnem-von-und-dassei-wiern-belbs-zu-ver-das-eweck-dir-ennen-sein

(Bense, Die Realität der Literatur 93–94, and see Bense and Harig qtd in Cory 6)

With the noise being removed from the signal and the Shannon approximations being applied, the garble resolves as the play progresses, mirroring the slow piecing together of what transpired; indeed, most of the play is relatively ‘classical’ as per Bense’s own terms. Nonetheless, the entire piece is a good representation of Bense’s theoretic and artistic modus operandi, displaying the reunification of logic and poetry in practice, where by logic it is understood the redundancy afforded by, for example, the rules and constraints of language (e.g., grammar or genre), and by poetry that which affords deviance from those rules. Though to be sure, deviance here is not to be regarded as a fault or a bug, but as generative. The poetic as noise and interference, as the aleatory (randomness and chance), as non-intelligibility, incites intelligibility. Intelligibility (or ordering, patterning) is a kind of living with, a writing with, noise. For Bense generative text explored through programmable machines making aesthetic choices became a way to reconcile the creativity and experimentation associated with art with the expected order of things associated with logic. For Bense, this was the new path for rationality, or what he called ‘productive intelligence’ (produktiven intelligenz) (Ungehorsam der Ideen 26). In this sense the algorithmic and concrete poetry experiments by the Stuttgart School were quite literally experiments at the border of unintelligibility and intelligibility, entropy and negentropy.

No doubt the cold media of machines and the incorporation of machine-generated language into the aesthetics of concrete poetry was, as we have suggested, a way of hosing down political passion, but it was not an outright celebration of technology either. There was a certain humility about the Stuttgart School project of opening communication between natural and artificial languages (a humility not always displayed across the Atlantic by the ambitious proponents of US AI; although, for example, Shannon and Marvin Minsky’s ‘useless machine’ was at least humorous). The critical rationalism displayed at Stuttgart, Ulm, and by the German experimental litterateurs in their respective circles, was a critique of the abuse of reason and rationality in that country’s recent past, a way of reclaiming, in fact, language itself. But having said this, Stuttgart certainly did not display the persistent, vocal critiques of the past and present as the returning Frankfurt School who passionately sought to uncover the societal contradictions present in any form of lingering positivism (even their own Marxist science). Others on a parallel track to the Stuttgart circle were more forthright in constituting an updated epistemology. One of Bense’s colleagues (located in Berlin, not Stuttgart) and fellow editor of GrStKG, Helmar Frank, called for what Friedrich Kittler would later name an ‘expulsion of the spirit from the humanities’ (Austreibung des Geistes aus den Geisteswissenschaften) (Kittler Austreibung; the German connotes exorcism). In his 1966 book on cybernetics, Frank proposed that by the ‘end of our century, … the ‘humanities’, which will by then be ‘modernised’, will be characterised by no longer being concerned with ‘spirit’ [‘Geist’] and its derivatives but rather ‘despiritualised’ into systems of information and information processing’ (113). Though certainly avoiding idealism and furthermore any vestige of Romantic subjectivity, the concrete poetry and the textual experimentation of the ‘projects’, as Bense called them, of generative aesthetics was less about de-spiritualisation than enabling a new, aesthetic configuration of intelligence and technics.

Generative aesthetics offers a critical counterpoint to the modus operandi of today’s generative AI. While both involve the artificial production of language, images, and other media, the philosophical underpinnings are significantly different. In the hands of its developers and investors, contemporary generative AI prioritises the automation of intellectual and/or creative labour (Pasquinelli). In this vision of machine intelligence, creativity is modelled with the goal to typically mimic or optimise, if not replace, existing human outputs. This vision promises labour efficiencies and lifestyle benefits at immense environmental cost (Crawford and Joler) and, increasingly it seems, psycho-social harm. Recent developments in agentic AI signal a further obscuration of the source and timing of decision-making across both commercial and social systems. In emphasising the programmatic and probabilistic nature of productive intelligence, Bense’s approach was, by contrast, not oriented toward mimicry nor even utility, but toward articulating machine and human languages to reunite the divided paths of logic and poetry. Indeed, as early as 1950 in his book Literaturmetaphysik (Literary Metaphysics) Bense defines this hybrid modality of the aesthetic object as ‘co-reality’ (Mitwirklichkeit). While today’s AI discourse frames itself as an inevitable acceleration into an imminent future, the Stuttgart approach to thinking with machines was an effort to reckon with the past, however obliquely such a task could be carried out given the School’s stated preference for experimentation. This was done to reconstruct forms of communication with an underlying cosmopolitan discourse that could speak to, and perhaps reboot, a damaged society, and in doing so reopen channels of thought that fascism and war had seemingly foreclosed forever. The opposite seems to be occurring now.

Whatever the case, historical consciousness is largely missing in today’s conversations about generative AI, whether critical or celebratory: AI is rarely treated as part of a wider, far more nuanced and transnational epistemic endeavour. In the case of West Germany, as one of its proponents, Friedrich Nake suggested, perhaps this was because the project of information aesthetics was ‘short-lived’ (Nake). Nonetheless, reintroducing Bense’s thought into mainstream AI history encourages us to see today’s generative AI systems less as evidence in a story of inevitable ‘progress’ but one in which a specific vision of machine intelligence gained dominance. It also should spur interest, for example, in revisiting the history of AI in the US to surface the importance of literature (and here I do not mean sci-fi fables) to the determining the form of its development.⁵

It would be tempting to understand the Stuttgart emphasis on information theory, early computing, and the randomisation of textual production as an exercise in the de-personalisation of creativity, a challenge to the notion of literature ‘expressing’ the spirit (Geist) of a person, nation, or an Age (for example, Hilder); a version, then, of the ‘death of the author’ argument being pursued at the same time across the border in France. This would explain, in part, the cool, formal nature of the new objectivity pursued by Bense and the Stuttgart School. But this is not the case. It was not that such literature was without subjectivity, that it was purely machinic, but that language should be seen in all its aspects and dimensions, human and machine, that is, semantically and syntactically, orderly and randomly, and indeed as national and transnational – all at once, now (jetzt).

The Stuttgart lesson is that language – the medium of intelligence, the medium of engagement – is a ‘concrete’ phenomenon. This is also how Bense’s attempt to repair Enlightenment rationality in the post-war context should be understood. Through a revised aesthetics, the concept of rationality should be extended to include uncertainty. For Bense and his circle the automation of creation was an experiment of balancing noise and deviance with ‘logic’. As such, noise, interference, and deviance were not factors to be erased to achieve successful performance but unavoidable dimensions of any communication. There is a humility to this position rarely found today. Finally, this productive, inclusive tension between logic and poetry was for Bense to be the basis for a new integration of the natural sciences and the humanities.

What Bense said in 1967 of the Tower of Babel once machine languages have joined it applies more than ever today: ‘our reflection, our consciousness, our communicative interest can ascend and descend [it] and’, he adds, ‘nothing has yet been decided about the further growth of this tower’ (‘Engagement und Experiment’). A more differentiated history of AI reveals research in machine intelligence need not been seen solely as a race to military, commercial, or epistemological singularity, but a way to encounter the ambivalences of rationality in an artificial world. With Bense and the Stuttgart School as guides, AI can be reconfigured as part of a project to redistribute intelligence to rebuild trust in both epistemology and politics. In this vision, a new set of coordinates for understanding AI and perhaps even deploying it becomes possible, and to this end, there are many other genealogies of AI yet to be explored. We should have a concrete approach to AI, not an abstract one.

Appendix

A friend of the

Friends of the intellectual dismantling of the ‘Ideology’

Friends of the intellectual dismantling of that German

Romanticism of civil, social and military

Prussianism

Friends of the intellectual dismantling of intolerance towards

liberal, democratic and Christian thinking

and institutions

Friends of the intellectual cosmopolitanism of art, science,

research, and literature

Friends of enduring reason and Enlightenment

Friends of universal and necessary world peace

based on a European world consciousness

A friend of the friends of the White Rose, the

Marble Cliffs and the Rhenish Group.

We only recognise as Germans those who distance themselves

from the lawlessness and crime of the

Ideology.

We recognise the legitimacy of the prosecution of

war criminals and their punishment

We want to constitute ourselves as Germans and Europeans

through the resolve

to punish our war criminals ourselves.

We want to carry out this punishment in the name of liberal,

democratic, Christian and humanistic law

in order to earn the right to rebuild German

research, German science, German

Universities and laboratories after political cleansing

and to make them effective for the

decisive world consciousness of the 20th century.

We want to carry out this punishment to show that

we distance ourselves from false provincial Germanness.

We ask for the friendship of the intellectuals of all nations.

The Friends of the White Rose, the Marble Cliffs and

the Rhenish Group.⁶

(Bense, Ausgewählte Schriften Vol. 4 527–28)