Reading time: 23 minutes
Translation by AB – November 5, 2022
Translation notes:
1) “Oulipo” is a very French thing and it doesn’t seem to have had many extensions outside France. We really hesitated to translate this article, but these few thoughts are finally more about “digital” itself, which knows neither borders nor ethnographic particularities. At least, apparently…
2) The word “potentialisable” does not exist in French; it is simply translated here “potentializable” which does not exist in English either. It could mean more or less “full of potentialities that can be revealed” and must be understood more or less ironically or playfully.
3) The verb “potentialiser” does exist in French and it is translated “to potentialize”, which does exist in English also. Usually, this verb is used to talk about the reinforcement of a phenomenon, such as the effect of a drug. This is more or less the same idea: “to potentialize” will mean here “to reveal full potentialities”.
Notice
This article was originally intended to be a simple ironic post-script to the previous article about these AIs that “talk” like us (GPT-3, LaMDA, Wu Dao… The blooming of “monster” AIs). But this little “Oulipian” postscript turned out to be much longer than expected and, above all, much more serious than ironic (one can joke with the Oulipo but not too much…). We have therefore made it a separate article.
The few allusions made here to “monster AIs” are therefore not explained and it is necessary to go back to the previous article if necessary (however, not doing so only moderately harms the understanding of the text).
Oulipo, short introduction
The “Ouvroir de littérature potentielle”, generally referred to by its acronym Oulipo, is a French group of inventive and innovative literature born in the middle of the 20th century. Its aim is to discover new potentialities of language and to modernize expression through writing games. The group is famous for its mathematical challenges imposed on language, forcing creative tricks. The Oulipo is based on the principle that the constraint provokes and incites the search for original solutions. It is necessary to thwart the habits to reach the novelty. Thus, the founding members liked to describe themselves as “rats who construct the labyrinth from which they plan to escape”.1
The French writer and Oulipian Georges Perec had thus proposed to get out of a labyrinth called a “lipogram”: his famous novel “La Disparition” was written without the letter “e” (the most frequent letter in French). This powerful constraint acts on the flow of the text like a “hand squeezing a garden hose”, to use an image proposed by the French writer Hervé Le Tellier, an active Oulipian and winner of the 2020 Goncourt Prize2. The text gains in strength, strain, and new potentialities. Many ways of “squeezing the hose” can be found on the Oulipo page devoted to the inventory of constraints developed over time3. For example, historian and writer Marcel Bénabou’s “definitional literature” consists in progressively extending a sentence by replacing the words with their definitions taken from a dictionary4. Thus, “Is Digital Potentializable?” (“Digital” used as a noun) can become “Is Digital full of potentialities that can be revealed” (our own definition), then, by applying the definitional constraint once again (online Merriam-Webster dictionary):
Is something (such as a device) characterized or operated by digital technology, full of abilities to develop or come into existence, be logically or axiologically able to be made known through divine inspiration?
In short, is Digital, like literature, Potentializable?
OuXpo
The Oulipo has never failed to meet every month since its creation in 1960 by Raymond Queneau (RQ) and the mathematician François Le Lionnais (FLL). The movement remains powerful and the “elements” of the wide OuXpo set, “X” variable subject, are sometimes invited to the “mardis de l’Oulipo”, these joyful public works proposed each month at the BnF in Paris (Bibliothèque nationale de France)5. Thus, music with the Oumupo (Ouvroir de musique potentielle – potential music) or theater with the Outrapo (Ouvroir de tragicomédie potentielle – potential tragicomedy) … These Ouvroirs are in a way “homomorphic” and all perform according to the same principle: in these workplaces, members try to make the potentialities of a subject X appear through formal and often mathematical constraints. These constraints act at the same time as tools, scalpels, scrapers… which, in a way, dismantle the subject and reveal the principles of its “action” behind the practices, habits and discourses. This is, in our opinion, the whole point of the OuXpian work: the potentialization acting as an unveiling is thus also presented, in the case of the Digital, and without exaggerating too much, as an ethical maneuver (Tristan Harris and the Swamp of Digital Ethics).
In spite of this compelling call of OuXpian practices, we strangely did not find any Ouvroir dedicated to the digital world and its different expressions. An Ou’inpo (Ouvroir d’informatique potentielle – potential computer science) was apparently baptized in 1997 but left no trace. An Ounupo (Ouvroir de numérisation potentielle – potential digitalization), an offshoot of the Piet Zwart Instituut, an art and design school in Rotterdam, held only one public meeting in 20186.
Is an OuXpian work on Digital then simply conceivable? We propose to explore here some of its conditions of possibility, starting with a long detour through… cookery! Because first we need to understand better, with a subject chosen (almost) at random, what OuXpian work in general consists of.
Analysis: Oucuipo
The foundations of this OuXpo devoted to cookery, an “Oucuipo” then (Ouvroir de cuisine potentielle – potential cookery), were laid in 1990 by writers Harry Mathews and Noël Arnaud. The tribute article “Cuisine potentielle en puissance : l’Oucuipo” written in 2012 by French food writer Bénédict Beaugé, begins with a simple and profound reminder (emphasis added)7 :
The connections between cookery and language have been underlined for a long time: Roland Barthes points out that, for Brillat-Savarin, the founding father of gastronomy, words and dishes present a certain equivalence, if not a sure equivalence. In “Pour une psycho-sociologie de l’alimentation contemporaine” Roland Barthes, again, shows that one can address cookery as semantics and that it is possible, then, to analyze it in terms of syntax or style.
In this short text, Roland Barthes writes (additions in brackets)8 :
What will the units thus identified be used for [ sweet, crispy…]? To piece systems together, syntaxes (“menus”) and styles (“diets”), no longer in an empirical way, but in a semantic way, so as to be able to compare them to each other: it is a question of revealing, not what is, but what means.
The OuXpian work seeks precisely to make this syntax/semantics articulation “creak”, to move syntax as one would change angle, to “see” how it goes about signifying. Reciprocally, semantics reveals by “edge effect” its tricks to pass into syntax. This is why a subject X is OuXpianisable if and only if, to use Beaugé’s words, it is “addressable as semantics” and “analysable in terms of syntax”. This will be, in any case, the starting axiom for questioning the potentialization of the Digital. This “axiom of potentialization” still needs to be clarified.
From semantics to syntax and vice versa
To oversimplify, we can say that the semantics are “encoded” in the syntax, the content in the shape. In a famous French radio program9, the gastronomical commentator conveys his knowledge and feelings (semantics) exclusively through language, intonations and various noises (syntax), which can then be encoded digitally or by frequency modulation. For example, here is a spoken syntax-recipe of corn bread encoded to be transmitted over the air10 :
We preheat the oven to 180°. We melt the butter, we add the sugar, we mix all that very well, then we add two eggs, one by one, then we mix the flour of corn with the baking soda, and then we are going to add also the buttermilk. Buttermilk is milk that is broken with a little vinegar, not fermented but has a slight taste like that. […] We put it all in the oven and pour it into a gratin dish. You get something halfway between a pastry and a cake. I personally love it, that little grainy texture. It’s a little crunchy, soft and sweet at the same time […]
Skillfully handled, the syntax-recipe restores with only words this culinary semantics evoking processes, flavors, aspects, textures…
Conversely, from syntax to semantics, the Oucuipian experiments. He fiddles the syntax by ingeniously applying constraints to it. One remembers the “Hundred Thousand Billion Poems” by RQ himself, the first Oulipian work respecting the constraint known as “cmmp” (In English then “htbp”). Georges Perec was inspired by this one to write “81 fiches de cuisine à l’usage des debutants”11 (“81 cooking cards for beginners”), a set of combinatorial recipes composed by permutations of three possible values of four “signifying units”: the main ingredient (sole, sweetbreads or rabbit), the cooking medium or method (pot, oven, pan), the final ingredient and the presentation. Here are two of his potential rabbit recipes:
Recipe #6 (rabbit à la bourguignonne): Generously spread 2 young rabbits with hot mustard. Place in medium oven for 40′, basting frequently. Deglaze with Noilly. Send a sauceboat of Burgundy sauce on the side.
Recipe #26 (rabbit with chipolatas): Generously spread 2 young rabbits with hot mustard. Put on high heat in a large frying pan, then lower the flame and let simmer. Deglaze with Noilly. Serve with chipolatas.
The principle is well understood. The recipe for Curnonsky’s sweetbreads (with broccoli) is then almost automatic:
Recipe #25 (Curnonsky’s sweetbreads): Finely slice 4 sweetbreads that you have previously soaked in lightly lemony water. Put on high heat in a large frying pan, then lower the flame and let simmer. Deglaze with Noilly. Serve with broccoli.
The Author
The potentialities of cookery are thus released by a double endeavor: 1) to forget the “semantics” of cooking i.e., gustatory, visual, and sonic sensations… and 2) to exercise constraints chosen a priori, drawing up a “labyrinth” from which the Oucuipian seeks to extract a legal gastronomic syntax. The sensations triggered by this new syntax are no longer the concern of the Oucuipians but of us (in that respect, the “rabbit with chipolatas” does not really appeal to us).
Finally, and above all, Georges Perec’s Oucuipian work reveals this by side effect (that we had discovered by other means here: GPT-3, LaMDA, Wu Dao… The blooming of “monster” AIs) : the recipe for rabbit with chipolatas obeys all the conventions of culinary syntax, which is enough to lend the apparent intention of the text (to cook a rabbit) to its “author”, real or imaginary. However, Perec’s algorithmic constraint cmmp does not carry any culinary intention, nor does any algorithm desire anything. FLL brilliantly summarizes this Oucuipian discovery, this horrible secret also known by the Oulipo12:
Any literary work is built from an inspiration (at least that’s what the author suggests).
And if there is no author (or less and less by the prodigy of the Digital) we believe nevertheless in the existence of inspirations. In other words, by habit, syntax always produces semantics.
Synthesis
A subject X is thus potentializable if it has a syntax that can obey a set of purely formal and often mathematizable constraints (permutations, repetitions, etc.). The syntax that we will qualify as “natural” corresponds to the encoding of the “natural” semantics of the subject, that is to say of its set of usual, conventional and real meanings. Potentialization comes, after more or less effort on the part of the OuXpian trying to escape from the constraints he has chosen, to add to the natural syntax an unlimited quantity of elements of “potential” syntax: recipes, poems, novels, musical pieces, etc. The syntax of subject X is thus widened with a considerable amount of material, which no other method could have created. These new experimental pieces of potential syntax are decoded in a semantics that we thus qualify as “potential” (for example, the disastrous impression caused by the recipe for rabbit with chipolatas).
A diagram is worth a thousand words:
Let the French writer Régine Detambel, herself quoted by Beaugé, put some “flesh” on this diagram13:
The constraint allows to take tracks which one would never have trodden: one reinvents lost words, conceives astonishing connections, reverses the clichés. And the more one is preoccupied with structural problems, the harder one works on one’s scaffolding, the more one expands one’s borders. And the territories that we conquer, we dare to roam them because we were looking elsewhere when we crossed our rickety bridges, when we came across our own monsters! We are right to say that we are not spontaneous when we try to be, because of these mountains of clichés which we are made of.
These “mountains of clichés which we are made of”, we have gathered them under the term “natural semantics”. In the Digital, they are immense and overwhelming.
The Digital as a technique of potentialization
Before facing the tough question of the potentialization of the Digital and of its avatars, we immediately note that the Digital can already serve as an instrument to potentialize (in any case, the Digital is the universal tool, the Swiss army knife of the contemporary world). It was already unstoppable for the mathematician FLL at the time, quoted again by Beaugé:
What some writers have introduced in their own way, with talent (even with genius) but some occasionally (forging of new words), others with predilection (counter-rhymes), others with insistence but in only one direction (“lettrisme”), the Ouvroir de Littérature Potentielle (Oulipo) intends to do it systematically and scientifically, and if need be, by resorting to the good offices of the information processing machines.
The “information processing machines” are even the best possible practitioners, because they can escape from hypercomplex logico-mathematical labyrinths. The Oulipo has been in close contact with machines and one can read on this subject the very interesting text of Camille Bloomfield and Hélène Campaignolle-Catel (“S+7”, “edges of sonnets” … are Oulipian constraints)14:
The digitization of corpus should allow, for example, the recognition of blank alexandrines in prose texts, or even a work on rhymes. With the help of the group’s contacts, various projects were underway: an automation of the S+n method (generalization of S+7), an attempt to computerize the “edges” of sonnets (carried out in 1965 on excerpts from Mallarmé, Leconte de Lisle, Hérédia, Baudelaire and Queneau), or the creation of a dictionary “numbering words by category (A = Adjective, S = Substantive, etc. )”, which “would allow, not only to realize much more quickly S+n (in particular for a very large n), but also to carry out this work by any machine”.
But it is one thing to turbinate the Oulipian constraints with machines and quite another to OuXpianize the Digital as such, question by us asked here.
Finally, the machine, however intelligent, can only serve as a factotum or a workhorse. For, no more than a monster AI spontaneously elaborates “prompts”, an information processing machine does not seem capable of inventing constraints. For that, but it is still only a hypothesis, the machine would have to participate in one way or another in natural semantics and be able to “feel what it can do” to move the syntax like this or like that. When Georges Perec chooses the constraint of the lipogram in “e”, firstly he knows that it will be difficult, but feasible and exciting, and secondly, he semantizes the constraint in a way. Indeed it is about disappearances in “La Disparition”15:
La Disparition is the novel of a disappearance which is the disappearance of the e, is thus at the same time the novel of what it tells and the story of the constraint which creates what is told.
The Oulipians would perhaps defend themselves, but a constraint is thus never totally free, i.e., it probably always carries a semantic ulterior motive: it is necessary to “feel what it can do” and to feel a malicious pleasure in it.
The monster AI, false friend of potentialization
After all, these AIs that automatically “respond” with a text, a piece of syntax, seem to participate in an Oulipian work of the prompt author. They indeed produce new texts from a minimal constraint: write a prompt and press “Enter”. However, an authentic Oulipian would not consider this constraint as admissible and therefore as “potentializing”. For if it is true that the user “feels what it can do” to propose such and such a prompt, there remain three prohibitive problems.
First, the AI only mimics the real syntax (the training corpus) with consequently a level of potentialization that is almost zero. Second, the same prompt produces different and unpredictable responses. However, there is a fundamental Oulipian principle that the writer Jean Lescure comments in his “Atlas of Potential Literature” as follows:
[…] the members of the OULIPO never hid their horror of the randomness […] « the OULIPO, it is the anti-chance », claimed one day without joking the Oulipian Claude Berge, what leaves no doubt on the aversion which one has for the dice.
This is because we must not misunderstand: potentiality is uncertain, but not hazardous. We know perfectly well what can happen, but we do not know if it will happen
We thus retain this constraint of constraints: a constraint is deterministic.
Thirdly, and perhaps most importantly, the pseudo-constraint of “producing a text using AI” deprives us of a challenge and of the pleasure of facing it.
Is Digital Potentializable?
The potentialization axiom implies that each OuXpo is homomorphic to all the others, in particular to the Oulipo, and thus respects the pattern drawn from our Oucuipian exploration (we only added “deterministic” to “potentialization with constraints”):
Our question is thus: can we have “X = something typically digital”, and are we entitled to found a potential “something typically digital” Ouvroir?
The problem which appears immediately, is that the Digital is not like literature, music or cookery: as a technology it carries neither syntax, nor semantics. We have sometimes qualified it as an “environment” (Emergence of a digital natural environment)16. So, an Oudipo (potential digital) would lack the most basic of materials to work with. But we can perhaps save the case by going in search of typically digital syntaxes (and by opening, why not, an Ouvroir dedicated to the research of Ouvroirs of digital potentialization…).
It is probably around “automation” that we should start looking, that is to say the more or less well-defined set of rules by which our actions and those of machines respond to and interpret each other (one of the simplest being the tuning of one’s sleep or physical activity with a smartphone). In a way, automation is like a digital “discourse” and the topics that can be drawn from it are quite clearly “addressable as semantics” and “analysable in terms of syntax”. But these topics are not simple to grasp and this difficulty reveals at the same time a bit of the “essence” of the Digital (this is finally where we wanted to go; see the conclusion).
Let’s illustrate this with two attempts to open up “something typically digital” potential.
Ounapo
When we navigate in Mundus Numericus, from mail to post, from post to page, from page to service, from service to document… we write a kind of sentence that obeys a grammar encoded in the structure of the net (links and algorithms). The natural semantics of the subject X = “navigation” contains what we want to “do” while navigating as well as our anticipations: search for information, calculate taxes, communicate with our network, pass the time, look for emotions… The Ounapo will incidentally disrupt this semantics by the side effect of deterministic constraints on the navigation syntax. It is a question of taking side roads, not by “randomness” but on the contrary, a marvel of the OuXpian method, in a perfectly controlled and deterministic way.
This Ounapian work would reveal, among other things: the “real” structure of the net, our habits and customs of navigation (these “mountains of clichés which we are made of”), the balance of power between our own will and that of the algorithms which direct our paths, unsuspected places, Our second natures…
At its origins, the Oulipo set two directions of work: a synthetic work, called “synthoulipism”, which consists in elaborating constraints and in experimenting with them; an analytical work, called “anoulipism”, which consists in searching in literature for examples of pre-Oulipian constraints and whose authors are called, rightly, “plagiarists by anticipation”. The Ounapo should in turn follow these two directions of work. By anounapism, look for existing constraints determining our navigation (and look for the motive, i.e., the semantics of their “authors”, in particular the digital firms). By synthounapism, elaborate new constraints, mathematical, joyful, absurd…
Synthounapism can be illustrated with this constraint-test that we call “navici” (let us specify however that an Ouvroir is always a collective; this constraint elaborated in solo is thus not valid). The navici consists in choosing any word, for example “Queneau” and navigating with it.
Take something to write down, write down the day, the hour and the chosen word (in Mundus Numericus, everything changes from one minute to the next):
Saturday, November 5, 2022, 19h23
Queneau
Open a browser, choose a search engine, note their names:
Brave DuckDuckGo
Start the search for the chosen word. A page of results is displayed:
The principle of the navici is to search for the first (or nth for an nth order navici) occurrence of the chosen word in the page (not “at sight” but by using the Ctrl-F search function) and to note the expression in which it appears. In this case (highlighted in orange):
Queneau
It must be recognized that so far not much has happened… But the navici continues like this: if the word appears in a hypertext link, note this link and click on it, otherwise look for the next occurrence of the word in the page (all occurrences appear in yellow):
Raymond Queneau | French author | Britannica
It is a link, so we click on it to arrive on the following page of the online Britannica encyclopedia (only the beginning is displayed here):
We look again for the first occurrence of the word in the page and we note the expression in which it appears (in orange):
Raymond Queneau
And so on. A text is thus assembled piece by piece, which depends only on the constraint and the structure of the web. The game stops when there is no more clickable link. The trace of this navici continues thus:
… Raymond Queneau, (born Feb. 21, 1903, Le Havre, France—died Oct. 25, 1976, Paris), French author who produced some of the most important prose and poetry of the mid-20th century. After working as a reporter for L’Intransigeant (1936–38), Queneau became a reader for the prestigious Encyclopédie de la Pléiade, a scholarly edition of past and present classical authors, and by 1955 was its director. From Queneau’s Surrealist period in the 1920s he retained a taste for verbal juggling, a tendency toward black humour, and a derisive posture toward authority.
But to really dismantle the syntaxes of navigation and to look for potential semantics, the Ounapians will have to be much more perverse, ironic, serious, obstinate, supportive… Magnificent constraints can certainly be invented to progressively unveil the action of the algorithms which write our “natural” navigations, which are so many practices, habits and dehumanized discourses.
Oupawpo
An Oupawpo is a workroom of potential web pages (Ouvroir de pages web potentielles), a workroom actually quite complex, a challenge to the imagination.
Web pages follow a specific syntax, both positional and aesthetic. This syntax obeys some well-known and internalized rules: menu bar at the top, legal and administrative notices at the bottom, organization by blocks, fonts, colors, etc. It responds of course to the “expected”, natural semantics of the Internet user (reading, listening, buying, searching…) but also – and here is the challenge – to intentional semantics, commercial or data collection for example, by facilitating reflexive reading and “natural” behavior. So, which syntax(es) should be used to reach this kind of double semantics?
This question can be illustrated with the Google search page that everyone knows and that looks like this:
The natural semantics of the Internet user-reader, decoded from this positional syntax, obviously “contains” the intention to carry out a search, but it also “contains” many other secondary sensations: familiarity, confidence, recollections… However, the underlying syntax is not this visual syntax but the “source code” of the page. Everyone can read it (right click on the page and “display the source code”). Here is a very short extract:
In other words, behind the clean Google page (“syntax 1”) lies a vast abstruse code (“syntax 2”). Which of these two syntaxes should Oupawpo tackle? The first one, from which the Internet user-reader decodes his own semantics? The second one, that encodes the intentional semantics of Google-author? Both agree on at least this: Google does intend to offer a search function and the user does intend to use it. But beyond this point of semantic agreement, the intentions differ…
The source code / syntax 2 of the Google page contains 140,000 signs, that is to say about sixty pages well packed. Nothing to do with the clean, almost empty page that is displayed. By comparison, the main page of this blog contains “only” 70,000 signs, that is to say half, in the service of a much denser 1 / visual syntax. An Oupawpian would then say that Google-author encodes much more intentions escaping the Internet-reader than Puissance & Raison17. The opacity of digital technology in general is due to the fact that the computer code, abstruse, complex, with a volume out of all proportion to its “appearances”, only responds for a minimal part to the natural semantics of the Internet user.
This observation is, however, valid for any technical artifact, which use is always necessarily simplified, designed to be rid of all the meanings that can disturb its simple use, i.e., simple consumption. Thus, the ease of piloting a car (its “reader-syntax” / syntax 1, pedals and dashboard) is out of all proportion to the complexity of its mechanisms (its “author-syntax” / syntax 2, mechanics and computers). However, digital technology has at least this essential difference: the computer code is not really a mechanism: it represents a mechanism. So, freed from material contingencies, the “hypertelias” of the source code linked to the author-intentions know in principle no physical limit: the author can write as much code as he wants, in particular code having no relation with the semantics – the expectations – of the reader: harvesting “data”, contextualizing the page to favor some reader-actions (click, buy…), etc.
In a different but still Oupawpian “style”, we know of at least one example of anoupawpism (thus of probable plagiarism by anticipation, even if we have not clearly identified the constraint at work): the home page of the online newspaper The Guardian, which comes from a source code of more than a million characters (the equivalent of an entire book is transmitted with each posting)! The syntax 2 of this page starts like this:
This syntax 2 is therefore not only intended for the algorithms that compute the Guardian’s reader-syntax but also for the engineers who design the source codes. These interlockings are the very matter and the real challenge of Oupawpo. So, we do have some presumptions of constraints but we let the Oupawpo roadmap be problematic. We just contribute with this preparatory and inaccurate scheme:
The hypothetical Oupawpo will undoubtedly find on its way the sentence “Code is law” of the American jurist Lawrence Lessig (Dimensions politiques de la blockchain – not yet translated) which can be translated as follows: the syntax 2 / source code makes the law.
Conclusion: about the “essence” of Digital
There are certainly other potentialisable digital subjects, i.e., “addressable as semantics” and “analysable in terms of syntax”, not to mention that Oupawpian complication concerning the divergent intentions of the author and the reader. So, there remains a mystery: in spite of all these possibilities, why is there still no digital Ouvroir homomorphic to the Oulipo? The inventory of the 6 possible arguments, resulting from our analyses and syntheses, is chilling.
Delivered raw and with “serony” (seriousness and irony):
- Digital is serious: it escapes any form of irony, and we sometimes wonder “how not to be afraid in front of this absence of reason devoid of any craziness?”18.
- Nobody has the time, the desire or the curiosity to potentialize the Digital. Even if Hervé Le Tellier suggests that “the majority of people are much more stupid than the average”19, this is not a reason to throw in the towel.
- Digital is perceived more or less consciously as an immanent technical medium, as un-OuXpianable as air. If you like, it is as if “the trees grew [here] in silence and [that] the animal kingdom limited its presence to obscure and mute act”20. No trees, no animals of course, but diffuse and uncontrollable digital presences…
- The syntactic-semantic relationship of the “reader” to the Digital is not of a tremendous richness since it is essentially a relationship of consumption. Therefore, human beings question very little the digital artifacts, “these objects so perfectly domesticated that they would have ended up believing them to have been created for their sole use”21.
- Digital is not meaning, it is number; and mathematicians know better than anyone that the mathematical syntax of number is not potentializable. For example, we say “I don’t know about you, but 14.28% of my life takes place on a Monday”22 but no one can replace 14.28% by another number without making a mess (not far from argument #1).
- The “readers” of digital syntaxes are mostly algorithms and machines, which makes us notice that the majority of today’s readers are much more cybernetic than average.
That’s mostly the problem.
1. ↑ (in French) Wikipédia – Oulipo
2. ↑ Wikipedia – The Anomaly
3. ↑ (in French)Oulipo.net – Contraintes
4. ↑ The origin of this constraint is explained by Marcel Bénabou himself here (in French): Oulipo 1966 : le tournant
5. ↑ BnF Les mardis de l’Oulipo
6. ↑ XPUB – March 28, 2018 – Special Issue 5 – OuNuPo
7. ↑ (in French) Bénédict Beaugé / Sociétés & Représentations 2012/2 (n° 34), pages 125 à 13 – 2012 – Cuisine potentielle en puissance : l’Oucuipo
8. ↑ Roland Barthes / Annales, 1961, 16-5, pp. 977-986 – 1961 – Pour une psycho-sociologie de l’alimentation contemporaine
9. ↑ « On va déguster » de François-Régis Gaudry sur France Inter.
10. ↑ François-Régis Gaudry / France Inter – November 23, 2014 – Happy Thanksgiving!
11. ↑ Georges Perec – 1985 – Penser / Classer
12. ↑ (in French) François Le Lionnais – 1973 – La LiPo (Le premier Manifeste)
13. ↑ Régine Detambel – La forme heureuse
14. ↑ (in French) Camille Bloomfield, Hélène Campaignolle-Catel – 2016 – Machines littéraires, machines numériques : l’Oulipo et l’informatique
15. ↑ (in French) Jacques Roubaud, quoted in Wikipédia – La Disparition
16. ↑ If we absolutely want to make it mean something, that is, if we want to find the human in it, it is at most as a tool. Here, we should go and see Bernard Stiegler, for example.
17. ↑ However, the Puissance & Raison site is “powered” by a WordPress code whose intentions we ourselves do not fully understand. 70,000 signs to encode the home page is far beyond our mere intention to post.
18. ↑ Raymond Queneau – 1941 – Les temps mêlés
19. ↑ Hervé Le Tellier – 1998 – Les amnésiques n’ont rien vécu d’inoubliable
20. ↑ Raymond Queneau – 1965 – Les fleurs bleues
21. ↑ Georges Perec – 1965 – Les choses
22. ↑ Hervé Le Tellier – 2003 – Guerre et plaies