Teaching a Rhizome How to Signify (A Work in Progress)

by Byrne Hollander

Table of Contents

Marc Ngui, the tracing should always be put back on the map


In this project I aim to demystify Gilles Deleuze and Felix Guattari’s rhizome as described in their 1980 book A Thousand Plateaus. I attempt to do so in a rather ambitious fashion–by applying their concept to this website as a whole; I treat this page as an assemblage to plug in seemingly disparate concepts. My primary goal is to provide a vision for what a hyperliteral rhizome might look like. D|G insist their rhizome is not a metaphor and simultaneously insist that any rhizome is too unstable to be represented, that “a rhizome is not amenable to any structural or generative model” (12). I suppose this project was born from my frustration with this notion that there is no way to visualize However, don’t feel obligated to engage with this post linearly. I hope readers find their own way through this page and its hyperlinks. In fact, here’s where I think things really start to heat up. I think this is the part where I take a detour to explain just what it is I’m trying to do here. One additional motivator for hyperlinks is the fact that (hopefully) this post will be read by people with a range of philosophical backgrounds; this will let people find their own entry point, let people who already understand trees skip over that, or those who are curious about deep learning can explore that topic first. this “image of thought”. My impulse was to consider dynamic, algorithmically generated graphs (with GUIs) as appropriate visualizations: the diagram would be in constant flux as new information enters the system and can be manipulated by users (e.g., filtering out certain information, zooming, etc.) However, this seems to be a poor starting point for this paper: I haven’t motivated the problem, set the stakes, or considered what it would mean for these diagrams to supposedly represent rhizomes. For instance Gregoriou argues “The goal is not to represent the rhizome but to implant it in thought. The effect they [Deleuze and Guattari] are after is not the understanding of the rhizome but a functioning, a whole apparatus that connects disparate, linguistic and non-linguistic things” (“Commencing the Rhizome” 240). Thus, I step back and propose a roadmap through which I hope to get to the crux of this problem.

But before I do, I first provide a little guidance for navigating this page. It’s filled with hyperlinks I’m particularly proud of the bibliography section (that is, its technical implementation)–anything that’s clickable opens a new window with the entire source document at your fingertips. A bunch of work went into this feature but I’ll spare you the technical details. I know I’m infringing copyright, but I implore you to suspend disbelief and imagine legality (and, for some, ethics) is not a concern here. I do it to prototype a digital rhizome. and you may have noticed that an Annotate button pops up anytime you highlight text. Consider them a recognition of the impossibility of creating a standalone work that explains a given concept once and for all. Plus, I want this page to be dynamic and to read people’s comments. I don’t think this instability implies any more chaos or disorganization than a typical academic paper. For instance, would willful disorganization have a table of contents? Additionally, D|G are of the same mind when it comes to any totalizing project:

We believe only in totalities that are peripheral. And if we discover such a totality alongside various separate parts, it is a whole of these particular parts but does not totalize them; it is a unity of all those particular parts but does not unify them; rather it is added to them as a new part fabricated separately.

While I would rather not paraphrase Deleuze and Guattari, this one is worth keeping in mind: when the sum of the parts become a whole, the whole itself becomes a part. One metaphor might be assembling an infinite puzzle: while you can connect pieces to form largest units, there will always pieces left to connect.

Which reminds me: while I have a laundry list of references and suggestions for further reading, this is not a rigorous academic paper. To be plain, I am playing fast and loose. I’m going for breadth rather than depth. I view D|G’s rhizome as a challenge; perhaps my lack of exhaustive analysis is not a cop out, but an application of rhizomatic thought? I don’t buy it. I intend to explain myself and represent my ideas as clearly as possible (which is why I’ve chosen this format and selected so many visual aids). I would rather be ambitious and fail than limit my scope to something uninteresting and/or unimportant. Thus, I take advantages of abstractions wherever possible. If someone has already explained a concept clearer or more succinctly than I have, why reproduce it when I could merely point you in its direction? Through this lens one might argue that thinking rhizomatically (if this can be considered ‘thinking rhizomatically’) is an excuse for laziness. Maybe. But if I’ve already found what I consider to be the answer to a particular problem, why should I put it in different terms? I view myself as a compositor See chapter 3 of The Language of New Media for more on selection and compositing. more than anything. If I’m to contribute to scholarship in any meaningful way, it won’t be for clever axioms, but for my ability to see how pieces Not only is this assignment’s deadline ominously looming, but I have another ambitious project due the following day. Which brings me to the advice I generally try to follow: “Never half-ass two things. Full ass one thing.” might fit together. I hope some interestingness comes of it. Moreover, this leads me to address a logistical issue: this project is primarily a submission for my English seminar. It could be best to consider this project as a kind of starting point for grappling with the rhizome. In some ways I am overtly essentialist and in other ways I am possibly mistaken. I’ve considered that this might not be the best way to discuss the rhizome, but it does seem like a novel approach. Further, it is dynamic and I can continue to refine it–it will not cease to exist for all intents and purposes once it is turned it (though I discuss the ephemerality of the web in a later place).

Further, the Annotate button is an implementation of Genius’s web annotator which lets you “add line-by-line annotations to any page on the Internet”. While there are other annotating solutions, no other service has a .js embed. I’m a big fan of Genius and I hope you consider creating an account and annotating something on this page. UPDATE: I removed this feature (for now)

Anyway, back to the roadmap. That was longwinded. I hope some took my advice to navigate this page non-linearly. As I said, this project began with a deceptively simple (or misleading) question: “What would a rhizome look like?” I now try to view the concepts on their own level and understand what problems the rhizome looked to solve. In other words, I try to unravel the context from which the concept sprang.

I begin by trying to locate the genesis of the rhizome project. I then argue that it differs from the root in that it tries to be ‘anti-Structure’. I explore what it means to think rhizomatically and consider the concept’s position as either metaphor or literal map. I find a mystical quality in the definition of the rhizome. I consider in what ways the Internet is supposedly not rhizomatic and if the maps rhizomes produce have those same characteristics.

I briefly consider the diagram. I mainly provide examples of past diagrams and propose a basic outline for diagramming a rhizome. I continue with a brief discussion of methods for collecting and classifying data that can be turned into maps of rhizomes. I conclude by considering this page itself as a rhizome. I provide an extensive bibliography.

The Rhizome as Image of Nomad Thought

Structure seems to be the overarching theme of A Thousand Plateaus. It is a book rife with geological and botanical metaphors, a book that seeks to explain space and relations within this space. It is a book of constant motion and instability. It is a book they claim is a rhizome with ‘plateaus’ in place of chapters. “Each plateau can be read starting anywhere and can be related to any other plateau” (22). Further, their rhizome “has neither beginning nor end, but always a middle (milieu) from which it grows and which it overspills” (21). I note that the rhizome is borrowed from botany, where it is a horizontal underground plant stem capable of producing the shoot and root systems of a new plant. Below is an illustration of the network structure that enables mushrooms to absorb nutrients from their environment. Richard Giblett’s Mycelium Rhizome, 2009

As Christy Wampole illustrates in Rootedness: The Ramifications of a Metaphor, botanical metaphors–especially those of the tree and root–have long been used to efficiently say what is otherwise extremely difficult or even impossible to communicate. Further motivating the importance of botanical terminology, Wapole notes critical theory’s reintroduction of the idea that “thought itself is botanical in nature and that humans should look toward plants as a model for dwelling in the world, because they provide an alternate paradigm for the treatment of space, time, consumption, and death.” Wampole, [Rootedness: The Ramifications of a Metaphor, 4-5.

Binary logic is the spiritual reality of the root-tree.

Thus, D|G consider the rhizome to be a radical enough shift away from the root to deserve its own concept. Reading suggestion: the introduction to A Thousand Plateaus provides a thorough comparison of the tree vs root vs rhizome in a different terminology. D|G also describe the six characteristics of a rhizome– I decide not to break them down one by one because I believe they function together and lack the same logic in isolation. While they admit “it’s not easy to see things in the middle, rather than looking down on them from above or up at them from below, or from left to right or right to left,” they argue it is the only way to see things. Rhizomatic writing is not written from a sedentary view: they assemble heterogeneity and do not seek fundamental unity. Thus, the question of where to begin seems strikingly lacking for nomadic writing. What can be our foundation if we begin in media res?

Where to begin in philosophy has always–rightly–been regarded as a very delicate problem, for beginning means eliminating all presuppositions. However, whereas in science one is confronted by objective presuppositions which axiomatic rigour can eliminate, presuppositions in philosophy are as much subjective as objective.

Yet, it is by confronting this important problem that Deleuze understands that there is no simple answer. Reading suggestion: Difference and Repetition’s Chapter III - The Image of Thought (though I cover some of it here)

Perhaps the start can never be the beginning– how can we tackle the issue of presuppositions? In Difference and Repetition, Deleuze argues that most of philosophy was founded on the notion of common sense as cogitatio natura universalis (good will of the thinker and good nature of thought). He calls this thinking the ‘pre-philosophical image of thought’ because it fails to question the nature of thought and being. This particular image of thought implies that “thought has an affinity with the true; it formally possesses the true and materially wants the true. It is in terms of this image that everybody knows and is presumed to know what it means to think.” This Image, according to Deleuze, “prejudges everything” and thus can be called a “dogmatic, orthodox or moral image”. In questioning the absoluteness of morality, Deleuze concludes that the conditions of a philosophy which would be without any kind of presuppositions are such: instead of being supported by the moral Image of thought, it would take as its point of departure a radical critique of this Image and the ‘postulates’ it implies. It would find its difference or its true beginning, not in an agreement with the pre-philosophical Image but in a rigorous struggle against this Image, which it would denounce as non-philosophical. As a result, it would discover its authentic repetition in a thought without Image” (131).

The ultimate distinction Deleuze finds between nonphilosophy and philosophy is that true philosophy lacks a foundation from which to build– there is neither truth in human nature nor in the world. Likewise, Derrida’s trace is rejected for having a ‘Structure’– as radical as it may be, Deleuze argues it still has meaning because it traces a preexisting structure or genealogical inheritance. While Wampole claims that “the rhizome is not radically different from the root. It is still an embedded form, its pace of growth is the same, and it represents yet another instance of thinking botanically about thought” (223), she doesn’t take this idea of structure particularly seriously. To her it is simply a nonbinary, nonhierarchical epistemology. She says that its lack of structure means its “thinking takes place in the ground” (223).

Haeckel’s Pedigree of Man. Tyranny in its purest form.

I think the important take away is that a graph of the root is iterating towards theoretically realizable edges (i.e., connections between concepts). Still it is a meaning not stable enough to subsist off. Likewise, it is not clear that concepts of different kinds can be on the same graph. Important to understand that the rhizome first signaled a break between Deleuze and Barthes / Derrida who believed in the “process of rewriting, reweaving, and grafting of all of the fragments and missing parts in a “Total Work or Magnum Opus” Lambert, In Search of a New Image of Thought, 52.. In other words, Barthes and Derrida’s philosophies imply that we are constantly in the process of delaying a totalizing image (at least while consciousness is out there). Still, it is about the structure of the world and what we are based off.


Still, D|G’s rhizome provides an ontology of relationships: everything is connected to everything else no matter how different they seem to be. I now make the trivial argument that some things can be more strongly connected than others. Take two slightly different sequences of DNA that end up expressing the same gene. They are more strongly connected (assuming that strength of connection is correlated with similarity) to each other than either sequence to any painting. Still, they are on the same graph.

Which brings me back to my initial question: How do you connect (i.e. create a map of) that which has no definitions between elements? I believe D|G respond: we don’t; the universe does. They tell us: “There exists a map, that much is for sure. Everything is on the map. Everything is connected.” Yet they claim we cannot manifest its structure. Before I go any further, I must motivate the notion that it deeply matters if we believe this statement. Further, the statement has different meanings depending on if we conceive of the rhizome as metaphor or literally. In the next section, I consider the implications of each position and relate them back to the Internet.

Rhizome as mystical or metaphor

Some critics have argued that A Thousand Plateaus should be recognized more for its poetics than for its philosophy: “by replacing argument with a flurry of pictures, [D|G] imagine they’ll change minds, since the human is such a visual creature. We no longer choose and commit to one picture; we accumulate and shuffle them” (Wapole 222). Wapole continues by arguing that this lack of argument implies a penchant for mysticism, which I hinted at in the previous paragraph. As a metaphor, this concept celebrates multiplicity and the dissolution of hierarchy and categories, and the horizontal proliferation of all aspects of life. These characteristics fundamentally influence how one experiences the world. I argue the rhizome that is an “image of thought” is a metaphorical rhizome. Lambert seems to agree: “Deleuze and Guattari’s rhizomatic experiment might be judged today as a failure, given that it was perceived as too abstract and appeared to convince no one, particularly not philosophers themselves” (60).

I now consider how metaphorical descriptions of the rhizome work. Some have argued it’s a style of thought more than anything. Wampole writes, “the rhizome’s vegetal matter does not last long enough to tell a tale and its surface is not dense enough to be inscribed with any lasting message. Any inscription on the rhizome would decompose along with it” (227). This statement relies entirely on the concept’s origin in botany– it doesn’t invoke any of the technical terminology from A Thousand Plateaus. Importantly, it argues that previous generations of a rhizome don’t mark later ones, largely because a rhizome is single-generational. The response that ‘it’s because the rhizome surface’s isn’t dense enough’ doesn’t have much logic to it– why can’t D|G’s rhizome have a dense enough surface?

Wampole continues by, as a tool, the rhizome can be used to mean whatever you want (here is a sample of titles of its various uses per chapter 7 footnote 2 of Rootedness):

“Becoming Rhizomatic Parents: Deleuze, Guattari and Disabled Babies” (Goodley, 2007); “Disciplinary Power, the Oligopticon and Rhizomatic Surveillance in Elite Sports Academies” (Manley et al., 2012); “The Rhizomatic Relations of A/r/tography” (Irwin et al., 2006); “Organizing and Regulating as Rhizomatic Lines: Bank Fraud and Auditing” (Bougen and Young, 2000); “A Rhizomatics of Hearing: Becoming Deaf in the Workplace and Other Affective Spaces of Hearing” (Crowley, 2010); “Rhizomatic Explorations in Curriculum” (Smitka, 2012); “Rhizomatic Thought in Nursing: An Alternative Path for the Development of the Discipline” (Holmes and Gastaldo, 2004); “COMMUNEcation: A Rhizomatic Tale of Participatory Technology, Postcoloniality and Professorial Community” (Broadfoot et al., 2010); “From Cluster F&%K to WTF: A Rhizomatic Reading of 4Cs Area Clusters in Technical and Professional Writing” (Smith, 2011); “Reuben’s Fall: A Rhizomatic Analysis of Moments of Disobedience in Kindergarten” (Leafgren, 2007); and “Rethinking ‘Foster Child’ and the Culture of Care: A Rhizomatic Inquiry into the Multiple Becomings of Foster Care Alumni” (Corcoran, 2012).)

If we choose to literalize the concept of the rhizome, we must grapple with how this concept is actually created. D|G themselves didn’t want the rhizome to be treated as a metaphor. Gregoriou explains that the rhizome must be rescued from its “literary over-coding in a pedagogical discourse as a metaphor for excessive multiplicity and radical openness. The rhizome must be made” (246). I agree wholeheartedly that the rhizome must be made.

When you hyperliteralize the rhizome, you lose all metaphor (of a certain kind) in exchange for a different kind of abstraction. We can now think what it might take to define the structure for such a map. In Protocol, Galloway argues that one major quality that disqualifies the Internet from being a rhizome is its protocol, i.e., that parts of the networks have designed systems for connecting and interacting with each other.

[Networks] are material technologies, sites of variable practices, actions, and movements. This is, perhaps, stated too strongly. Yes, metaphors do materialize and corporealize, and, in some sense, metaphor is consonant with language itself. But discussions of networks–especially in cultural theory–have too often slipped into “vapor theory,” eliding a specific consideration of the material substrate and infrastructure with a general discussion of links, webs, and globalized connectivity.

The infrastructure is of the highest importance here. Once we have protocols, a true map from content to content is created. We must additionally create a scheme for placing value on the edges between nodes and then do something interesting with this graph, but protocols must be the first step.

Protocols do not perform any interpretation themselves; that is, they encapsulate information inside various wrappers, while remaining relatively indifferent to the content of information contained within. The consequences of this are legion. It means that protocological analysis must focus not on the sciences of meaning (representation /interpretation / reading), but rather on the sciences of possibility (physics or logic).

I think my project really intensifies around here.By positing the existence of a literal rhizome (and thus a map), D|G implicitly argue that a protocol exists for adding anything (“semiotic chain(s)”) to the graph and a function exists to evaluate the relationship between anything and anything else. While they insist that “[u]nlike a structure, which is defined by a set of points and positions, with binary relations between the points and biunivocal relationships between the positions, the rhizome is made only of lines: lines of segmentarity and stratification as its dimensions, and the line of flight or deterritorialization as the maximum dimension after which the multiplicity undergoes metamorphosis, changes in nature” and that “[t]hese lines, or lineaments, should not be confused with lineages of the arborescent type, which are merely localizable linkages between points and positions,” the six characteristics they use to define a rhizome do not support all of their assertions (22). They say everything is connected, though by lines and not lineages. But really, this definition of a rhizome is meaningless in my pursuit of visualizing a rhizome: they’ve already said every “rhizome pertains to a map that must be produced, constructed” (21). If every rhizome must have a corresponding graph, I will focus my energy there as graphs are an undeniably more stable concept than that of the rhizome. Graph terminology I will use from here on out:

  • Every node x is anything that could be in a conceivable rhizome r.
  • Every node y is anything that could be in r and is not x (no duplicates in the rhizome).
  • I call an edge anything that links node x to node y in r.
  • Edge weight refers to the strength of a connection (in units – it’s only rhizomes that are “composed not of units but of dimensions, or rather directions in motion” (21)).
  • Each edge’s weight is calculated by executing function f on the start and goal nodes (“weight_xy = f (x, y)”)
  • There is also a function fp that calculates the cost of traversing path p to get from start to goal (“cost_xyp = fp (x, y, p)”)
    • This function exists so we can cluster nodes with strong connections.
  • All my edges are directed (i.e., the edge from x to y might not have the weight as the edge from y to x).
  • A ‘line of flight’ is a vector or sequence of vectors that make the edge weights of a rhizome unstable

While such a rhizome r must exist if we are to literalize the concept itself, we lack the protcol to create nodes and assign edge weights and path costs. Nevertheless, a protcol and algorithm must exist such that the r is constructed. Still, if there exists a map, there must exist a method to compute it. D|G might argue that various lines destabilize the map, but this map doesn’t seem particularly anti-genealogical What Deleuze and Guattari call the principles of cartography and decalcomania repeats the argument of the notion of a plane of consistency that is not amenable to any structural or generative model. BUT maps can be genealogical. In one generation, we can add, remove, or reweight vertices. Their likely argument is that because context has changed, so has the concept(s). While everything might have reprocussions on everything else (and does, according to them), that doesn’t mean they are fundamentally uncomparable between generations. If everything exists on a continuum, why can’t things merely shift rather than be destroyed? The result is still a map that doesn’t find the meaning, but creates it.: we can create arbitrary timesteps and see how the map changes from step to step. Moreover, while we do not know the details of how the map is generated, the universe’s algorithm always returns an answer. (If the map were not computable, how could relationships between concepts exist?)

I now argue that protcols, functions for assigning edge weights and path costs, and analysis of the resulting diagram must exist for all digital information. Note that this is exactly how Galloway argues the Internet is not rhizomatic– “the Internet works too well. If the Internet were truly rhizomatic, it would resist identification. It would resist the deep, meaningful uses that people make of it everyday. The Net is not narrative-based, or time-based. But it still enthralls users, dragging them in, as television and cinema did before it. How? The answer is in the form” (64). It sounds like we do want identification (note that the internet is still a ‘nonsignifying system’) if it gives us the ability to make the network have deep, meaningful uses. And as shown above, a rhizome by its nature must be identified to exist; to reiterate, we might not be able to idenitfy it, but it has an identity. Every rhizome is a map is a network.

Protocol not only installs control into a terrain that on its surface appears actively to resist it, but in fact goes further to create the mostly highly controlled mass media hitherto known.

Why am I so confident that we can create a protocol for anything digital? Because everything digital shares a universal encoding in binary. If there exists an assemblage where everything on it is represented in binary, we can place them in a context where they relate to one another based on their relative position and their contents. The structure of the internet is worth considering. Currently, most of the web (this site included) is written based mainly on documents written in Hypertext Markup Language (HTML). Using metadata tags, computers can categorize the contents of the web pages. In addition, Semantic Web solutions enable structured collections of information such that computers can conduct automated reasoning. However, as Berners-Lee et al. note in the article, artificial intelligence researchers have been trying to structure information so reasoning can be computed on it. Likewise, AI researchers have come up with methods to conduct automated reasoning.

To understand how we can compare binary code to other binary code, we must view each file on a continuum from bits to databases. It is impossible to know where one concept begins and another one ends (especially when we lack semantic networks), so we must consider each expressing section of code both separately and as a whole. While I won’t propose an algorithm for comparing content within and between data types, it certainly seems plausible. There are other ways of digitally mapping connection. One obvious example is HTML itself– Google’s PageRank algorithm crawls the web in order to determine which sites are most strongly connected to a given query.

Diagrams meaningfully spatialize relationships

I debated removing this section, like I have so many others in the process of making this page. But I think the visualizations are pretty and there are enough external sources writing interesting things about the diagram as it could be applied to the rhizome, so I’ve decided to keep it in here. You can see how the project has evolved over time by comparing commits on Github.

What distinguishes the map from the tracing is that it is entirely oriented toward an experimentation in contact with the real. The map does not reproduce an unconscious closed in upon itself; it constructs the unconscious.

We don’t get a top-down view when creating a map: a rhizome isn’t vertical enough for it to be possible. But a visualization can still be valuable. I think the most important part of any network visualization is the ability for the user to interact with it– filter out data I really think filtering out information is vital. While we lose context, it is impossible to make sense of an overload of data. Data becomes much more manageable when you can seek out and manipulate particular characteristics., zoom in and out, etc.

What are maps for? This is simple enough to answer: to find one’s direction, to orient movement toward something.

While much thought has gone into visualizing data Edward Tufte has done a bunch of work on data visualization – in fact, this site’s design was inspited by his books. Here’s his 2001 book The Visual Display of Quantitative Information, I find that many visualizations are beautiful but impart very little information. While I highly recommend the graphics in Lima’s Visual Complexity: Mapping Patterns of Information and think they are beautiful, I can’t help but wonder who they mean things to? Still, Lima provides a vision for data visualization moving forward:

An early GUI for network analysis.

This new cybernetic process will rely on our exponential increase in computational power to process and filter the massive data required to provide a real-time documentation of global activity and resulting behavioral patterns. Each element within our collective biosphere will need to be carefully tracked, interrelating multiple factors such as geologic and climatic conditions, predatory traits, and life cycles. In this complex and multidimensional data environment, the role of visualization will be key in pro- viding the capacity to recognize the emergent patterns and processes of these phenomena. Visualization will itself become organic, as it will need to adapt to simulate information from a wide spectrum of sources, ranging from micro/organic to macro/planetary states. The role of artificial intelligence will be critical in creating this new cybernetic form of resistance, revealing abnormal trends and anomalies and giving us the ability to utilize resources.

One somewhat interesting interactive visualization of the web

An interactive network visualization. Go drag stuff around!

Simulating and Classifying

This section is mostly to provide starting points for how we can think about getting and classifying data to be entered into a rhizome map. I go into very little detail, but there are several relevant papers in the references and relevant literature.

I return to the AI project of classifying data. One way to classify data is with mass assignment based neuro-fuzzy networks. This type of neural network has several relevant qualities: it is dynamic (its input does not need to be of a fixed size), the number of categories to classify things into does not need to be fixed, and assignment to categories can be continuous (as opposed to binary). While imperfect, the important thing is that there are ways to methodically relate digital forms to one another. We might be interested in using a semantic neural network with fuzzy values if our rhizome is all text or a deep convolutional neural network if we only have images. The main argument here is that we can try to approximate the universe’s map making functions.

Another interesting idea is using a rhizome’s data to point us towards what might happen. While somewhat of a longshot, RNNs can dream– you train them and then they invent (usually nonsense, but well formatted nonsense) based on what they’ve seen. There’s a great blog post by Andrej Karpathy called The Unreasonable Effectiveness of Recurrent Neural Networks.

Each rectangle is a vector and arrows represent functions (e.g. matrix multiply). Input vectors are in red, output vectors are in blue and green vectors hold the RNN’s state (more on this soon). From left to right: (1) Vanilla mode of processing without RNN, from fixed-sized input to fixed-sized output (e.g. image classification). (2) Sequence output (e.g. image captioning takes an image and outputs a sentence of words). (3) Sequence input (e.g. sentiment analysis where a given sentence is classified as expressing positive or negative sentiment). (4) Sequence input and sequence output (e.g. Machine Translation: an RNN reads a sentence in English and then outputs a sentence in French). (5) Synced sequence input and output (e.g. video classification where we wish to label each frame of the video). Notice that in every case are no pre-specified constraints on the lengths sequences because the recurrent transformation (green) is fixed and can be applied as many times as we like.

With this we have the tools to generate new concepts– it reminds me of the infinite monkey theorem which states that a monkey hitting keys at random on a typewriter keyboard for an infinite amount of time will almost surely type a given text, such as the complete works of William Shakespeare. But what if instead of typewriters, they had an interface for selecting context-sensitive text recommendations. For example, if a recurrent neural network were trained on Shakespeare’s complete works, it could generate new text in his style. Because the monkeys are choosing phrases that Shakespeare may have used, the speed with which they would likely create a great new play in Shakespeare’s style would be significantly faster than if they were using a standard typewriter. I think this same idea can be applied to generate new concepts and add them to a given rhizome.

Next, I mentioned above that we can only algorithmically construct maps on the digital. But everything analog can be represented digitally without a loss of information. Although we can emulate analog processes to any desired degree of accuracy with digital computation, we lose several orders of magnitude of efficiency in doing so. A single transistor can multiply two values represented as analog levels; doing so with digital circuits requires thousands of transistors. See Carver A. Mead’s “Neuromorphic Electronic Systems” So how can we get a digital representation of our universe? We simulate it. Some prominent people think it’s very likely that our own universe is a simulation. And that soon enough we’ll be simulating our own universes.

This post as a rhizome

While unwieldly, I managed to create this assemblage “on a single page, the same sheet: lived events, historical determinations, concepts, individuals, groups, social formations.” Ok, I don’t have ‘social formations’ or ‘individuals’ on this page, but I try to be pretty heterogeneous in my content: the site has a video, static images, an animated image, embedded code, javascript, HTML, and links to HTML pages embedded with pdfs and ePubs I really wanted to be able to embed the documents directly in the site, but it was taking too long to write the necessary code. Plus it was slow and might not work on all browsers.. In terms of kinds of content, I think this page makes A Thousand Plateaus look homogenous by comparison. Which begs the question: what if D|G had the technology to enable radical interconnectedness? It also reminds me of Ted Nelson’s xanalogical structure which would enable parallel reading of a source and the location it is cited. I was frustrated reading their book when they assumed the reader has expert-level familiarity I think diagrams would greatly help clarify their writing. Marc Ngui has a project where he is illustrating ATP. I have some of his drawings embedded on this site.

I think one major innovation this post has that ATP lacks is its invitation to collaborate: the Annotate cues enable anyone who reads this to ask questions, add information, or challenge something. Further, anyone can easily clone my Github repository for this project and emulate this post’s style (or slightly change it and host it on their own domain). Looking forward, I think a better version of this would be dynamically hosted (Github only does free static hosting) where each reader can customize the page and see other people’s customizations. I also think the hyperlinks and citation linking mean this work is much more navigable than ATP.

While this marks the end of the project, I intend to expand it once the rest of my finals are over. There’s a lot left to say, but I hope I’ve created my own kind of map for getting there.



Berners-Lee, Tim et al. “The Semantic Web”. 2001.

Burnett, Kathleen. “Toward a Theory of Hypertextual Design.” Postmodern Culture 3.2 (1993). Web.

Caliusco and Stegmayer. “Semantic Web Technologies and Artificial Neural Networks for Intelligent Web Knowledge Source Discovery”. 2010.

DeLanda. Deleuze: History and Science. 2010.

Deleuze, Gilles. Difference and Repetition New York: Columbia UP, 1968. Print.

—. “Postscript on the Societies of Control”. 1992

Deleuze, Gilles and Guattari, Felix. A Thousand Plateaus. 1980.

—. Anti-Oedipus. 1972.

Dronsfield, Jonathan. “Deleuze and the Image of Thought”. 2012.

Drucker, Johanna. Graphesis: Visual Forms of Knowledge Production. Cambridge: Harvard UP, 2014. Print.

Galloway. Protocol: How Control Exists after Decentralization. 2004.

—. “What Is New Media?”. 2011.

Gregoriou, Zelia. “Commencing the Rhizome: Towards a minor philosophy of education”. 2004.

Humphreys, Chloe. “Rhizomatic Writing and Pedagogy- Deleuze & Guattari and Heidegger”. 2013.

Koh, Chuen-Ferng. Internet: Towards a Holistic Ontology. 1997. Web.

Krizhevsky, et al. “ImageNet Classification with Deep Convolutional Neural Networks”. 2012

Lambert, Gregg. In Search of a New Image of Thought: Gilles Deleuze and Philosophical Expressionism. Minnesota: U of Minnesota, 2012. Print.

Larkin and Simon. “Why a Diagram is (Sometimes) Worth Ten Thousand Words”. 1987.

Lima, Manuel. Visual Complexity: Mapping Patterns of Information. 2011.

Manovich, Lev. The Language of New Media. 2001.

Moktefi, A. and Shin S. Visual Reasoning with Diagrams. 2013.

Pinker, Stephen. “A Theory of Graph Comprehension.”.Artificial Intelligence and the Future of Testing. Ed. R. Feedle. New Jersey: Erlbaum Hillsdale, 1990. 73-126. Print.

Roverso, Davide. “Soft computing tools for transient classification”. 2000.

Schmidhuber, Jurgen. “Deep Learning in Neural Networks”. 2014.

Semetsky, Inna. “Deleuze’s New Image of Thought, or Dewey Revisited”. 2003.

Shin, Sun-Joo. The Mystery of Deduction and Diagrammatic Aspects of Representation. 2015.

Sokal, Alan and Bricmont, Jean. Fashionable Nonsense: Postmodern Intellectuals’ Abuse of Science. 1997.

Veltman, Kim. “Towards a semantic web for culture”. 2001.

Wampole, Christy. Rootedness: The Ramifications of a Metaphor. Chicago: U of Chicago, 2016. Print.

Wang, et al. “Generalized Autoencoder: A Neural Network Framework for Dimensionality Reduction”. (2014)

—. “Semantic Clustering for Text Categorization”. 2015.

Zdebik, Jakub. Deleuze and the Diagram. 2012.

Relevant Literature

Not yet organized, but here My friend joked that this phrase could be the title of his memoir. For that reason alone, I’m keeping this section as is.