Artifices of Intelligence/Intelligence of Artifices

example cellular automata

The following is the text of a talk I gave on Derrida’s La Vie La Mort Seminar in February 2020 at a conference in Ann Arbor organized by Sergio Villalobos-Ruminott.

Thank you all for being here; I would especially like to thank Sergio for the invitation to speak. I have to admit that I am both honored and a little bit nervous about being the first speaker, perhaps it is not the case but I feel a certain responsibility to set a tone for our conference. While I hope to do my best in this regard, I have to say I am even a little more nervous about my abilities to do so given my present circumstances. When I began planning for this talk a few months ago, I set myself what appeared to me a rather ambitious research question, namely, to ask what are we to make of this question of life-death and the program in the light of advances and excitement (likely over-excitement) around artificial intelligence. But, and apologies for sharing such personal details, in the intervening weeks I have had to deal with a family medical situation in which our child has been in and out of the hospital dealing with frequent seizures. So, I hope you will accept my apologies that my talk will have to be more speculative, less well cited, fragmented and reframe some of the work that is present in my book on deconstruction and cyberwar rather than rigorously pursuing new research. Yet, I will still attempt to at least chart for you all where I see this research going and some of the questions I believe it raises in the context of Derrida’s meditations on life-death.

Moreover, though I fear I can do little with these thoughts at the moment, on a personal level I have been thinking deeply of the questions raised by this seminar and the work many of you in this room have done around its questions, as I have been searching personally for metaphors for these organic neurons, that may or may not bear similarities to those artificial or computational neurons that have become the workhorses of modern artificial intelligence, and their networks, energetics, discharges, stimulations and ultimately storms, seizures and related malfunctions, that is their ceasing to function or their functioning differently. And it has struck me that there are many questions to ask about this economy of energetics, and its différances, alongside Derrida’s writing on Freud’s attempts to describe these systems, and ultimately the extent to which these energetic economies are best understood as a form of arche-writing or perhaps more clearly as that arche-violence with which Derrida also offers to name the same movement of différance.

I would like to start with four epigraphs of sorts; each related to attempts to produce or reproduce life or intelligence computationally, which at our present moment amounts to producing it in silicon.

A complete discussion of automata can be obtained only by taking a broader view of these things and considering automata which can have outputs something like themselves. Now, one has to be careful what one means by this. There is no question of producing matter out of nothing. Rather, one imagines automata which can modify objects similar to themselves, or effect syntheses by picking up parts and putting them together, or take synthesized entities apart. In order to discuss these things, one has to imagine a forma set-up like this. Draw up a list of unambiguously defined elementary parts. Imagine that there is a practically unlimited supply of these parts floating around in a large container. One can then imagine an automaton functioning in the following manner: It also is floating around in this medium; its essentially activity is to pick up parts and put them together, or, if aggregates of parts are found, to take them apart.
This is an axiomatically shortened and simplified description of what an organism does. It’s true that this view has certain limitations but they are not fundamentally different from the inherent limitations of the axiomatic method. Any result one might reach in this manner will depend quite essentially on how one has chosen to define the elementary parts.
– John von Neumann, “Re-evaluation of the problems of complicated automata—problems of hierarchy and evolution”, 1949

We consider Conway’s latest brainchild, a fantastic solitaire pastime he calls ‘life’. Because of its analogies with the rise, fall and alternations of a society of living organisms, it belongs to a growing class of what are called ‘simulation games’–games that resemble real-life processes…The basic idea is to start with a simple configuration of counters (organisms), one to a cell, then observe how it changes as you apply Conway’s “genetic laws” for births, deaths, and survivals…You will now have the first generation in the life history of your initial pattern…Mistakes are very easy to make, particularly when first playing the game. After playing it for a while you will gradually make fewer mistakes, but even experienced players must exercise great care in checking every new generation.
– Martin Gardner, “Mathematical Games: The fantastic combinations of John Conway’s new solitaire game ‘life’”, 1970.

It seems to me that today artificial intelligence represents the fourth narcissistic blow to humanity. Recall Freud’s famous statement [recounting the historical blows dealt by the sciences to “the naive self-love of men”] First Copernicus, followed by Darwin, then psychoanalysis, and now the fourth blow: the capturing of intelligence by its own simulation, exceeding and transcending it.
– Catherine Malabou, Morphing Intelligence, 2018

If a technology tested in Europe proves to be too opaque or fails to comply with the rules in place, regulators may order the company to reboot the AI and make it learn from scratch based on different data
– Valentina Pop, Big Tech to Face More Requirements in Europe on Data Sharing, AI, Wall Street Journal, February 19, 2020

example cellular automata

If the, or at least one, question posed in La Vie La Mort could be summarized as: how are we to confront this moment when in the history of science the concepts of writing, of trace, of différance, become necessary to advance a scientific project that still clings to metaphysics, especially a metaphysical conception of life, what I hope to do today is to open a corollary question or perhaps the inverse of this question: namely what happens to metaphysics when life, and its products, such as reproduction and intelligence, can be expressed as a program. Alongside the question of understanding life as program, we confront the question of the program and the computer being able to “give birth” to life, to a form of life, of life-death, that if we take it seriously immediately calls into question the metaphysical notion of life, of intelligence, rationality and the human.

To start from the last example from these epigraphs, of artificial intelligence, I will admit that up until the point I sat down to write this paper I fear I have not taken it seriously enough. I have tended in many ways to dismiss it variously as dimensionality reduction, statistics without rigorous hypothesis testing, unable to pose or even answer actually interesting questions etc. etc. But, I realize now perhaps these have all been defenses against including advances of artificial intelligence within the list of Freud’s humiliations of man as Catherine Malabou does. Perhaps it is as such a humiliation that it could shake our very idea of intelligence, and especially its relationship to humanity and a metaphysical understanding of life. My intention in this talk is to focus on this question of artificial intelligence, that is an intelligence that is artificial in so much as it is ‘programmed’ and its potential affront to metaphysical concepts of life, intelligence, etc. as well as its simultaneous threat to reify metaphysics to the nth degree and then to return to this question of the program especially as it is presented in Derrida’s later work.

To begin, it is important to point out as Malabou does and many others along with her the long racist and eugenic history and nature of many discourses around intelligence, associating intelligence with what Sylvia Wynter calls the “overrepresentation” the human, that is with its white western ableist and male conception. While this can be well traced through the history and discourses of IQ testing, it can also be seen it at work in the infamous Turing, in which true artificial intelligence is measured not by some deep understanding that always appears ineffable when one tries to determine its existence, but rather by the ability of a computer to fool a human judge into thinking that it too is a human.

Thus, we should ask the question of which humans would be able to pass this test, that ultimately requires a certain acumen in the language of the examiner. Moreover, this use of language often tends to mean not so much the highest expression of rationality as the enlightenment dreamt it but rather a certain default of sense and of reason that marks the human use of language. It would not be difficult to show that this concept of intelligence, both as represented in the IQ test and the Turing test, is thoroughly metaphysical and anthropocentric insisting that “man” is both the possessor and measure of intelligence.

To take seriously the advancements of artificial intelligence, especially as a humiliation to man, must involve in a sense abandoning or at least thinking beyond such a conception. At the same time the discourse around such developments, especially those that adhere to certain ideas of progress, reveal the deepest dreams of metaphysics, specifically those of Hegel that Derrida cites in the seminar, where life, or here intelligence, should in progressing transcend the individual and with it death.

In this way, we touch on a more general concern that is at work in computation and programming in general, especially as it simulates and produces the products of life and the sciences writ large, namely their simultaneous allegiance and foundation in a logocentric metaphysics and their threat to outpace that metaphysics. Derrida starts the Grammatology stating that logocentricism controls in one and the same order, amongst other things, “the concept of science or the scientificity of science—what has always been determined as logic—a concept that has always been a philosophical concept, even if the practice of science has constantly challenged its imperialism of the logos, by invoking, for example, from the beginning and ever increasingly, nonphonetic writing.” Likewise, Francesco Vitale says of Derrida’s work in the seminar that he “he verifies the deconstructive effects of the recourse to the theory of information and in particular to the notions of “programme” and “writing” in the context of the life sciences. It is worth remarking, however, that, from the first pages, Derrida seems to be very suspicious about the alleged emancipation of biology from philosophy.” I think this is ultimately what is at stake both in biology and the simulation and rearticulation of life and intelligence in computer science. In short the turn to arche-writing in the biological sciences at stake in la vie la mort and the more recent turn to programming, statistics and networks in opaque machine learning algorithms threaten in a sense simultaneously to cement and to oust metaphysics.

In this regard, I have been struck for a number of years now by a single sentence about cybernetics, in the Grammatology; Derrida says in the form of a conditional:

“If the theory of cybernetics is by itself to oust all metaphysical concepts — including the concepts of soul, of life, of value, of choice, of memory — which until recently served to separate the machine from man, it must conserve the notion of writing, trace, written mark, gramme or grapheme, until its own historico-metaphysical character is also given up.”

I have chosen to translate this last word as “give up”, especially in the sense of turning oneself in to the authorities, rather than “expose” or “denounce” as Spivak has, for a number of reasons, that I should like to make clear to you.

This giving itself up suggests that what is at stake is simultaneously admitting what has taken place, abandoning a former way of being, that is saying to oneself “I can’t go on this way anymore; life on the lam is no longer possible” and ultimately taking responsibility. But let us be clear this taking responsibility does not necessarily mean in a personal or moral sense but rather in the face of a situation, and here an interpretation gets especially perilous, but perhaps what cybernetics gives itself up to is even a certain authority, which here may be metaphysics qua science itself; we imagine then in a way the police, who have always in a way been synonymous with metaphysics, turning themselves in to themselves. When one gives one self up in this way at minimum they only admit that they can no longer avoid the consequences of what has happened or at least what they have been accused of regardless of whether or not one feels the law has actually been broken or that the law is just. We of course here risk the use of a metaphysical accounting for this abandonment, but that is precisely what is at stake in this sentence is metaphysics qua cybernetics giving itself up, exposing and denouncing, abandoning its life on the lam.

Moreover, what strikes me about this sentence is that Derrida says that cybernetics would have to preserve writing, trace, etc. up until that moment it gives itself up. Remembering that this is all under a conditional “if”, were cybernetics capable of fulfilling this conditional in the affirmative, on the other side of giving itself up, it could even perhaps go further than Derrida and do away with these quasi-metaphysical concepts that teeter on the edge between metaphysics and some elsewhere, but still trapped in its enclosure. What it seems to me Derrida is suggesting here is the possibility that cybernetics could possibly escape metaphysics, that it bears within it the possibility of a non-metaphysical science. We may even suggest the hypothesis that what ultimately lead to such rapid abandonment and collapse of cybernetics at the forefront of science was that it drew in its later iterations too close to this precipice, but such hypotheses must be held in reserve for the time being.

To continue on this path of inquiring into whether the advancement of “the program” as a deconstructive force would allow cybernetics or the biological sciences to outpace metaphysics, recognizing full well that it makes no promises to do so, requires than accounting for the function of such a program. I have taken this question of the program up at length in my book, so I will not belabor the point, but it should be noted in this regard that at points especially I think in his later work and inheriting in some ways Heidegger’s antipathy to cybernetics and the program, Derrida insists that the program in its calculability forecloses the event and the possibility of that which arrives. For him, the unconditionality of the openness to what arrives under the name of deconstruction marks “the end of the horizon, of teleology, the calculable program, foresight, and providence.” That is to say that program is too machinic, to determined, to predictable and thus in a way captured within the enclosure of metaphysics

The programme is dual for Derrida: In the seminar he states:

“The heterogeneity of causes and effects, the non-deliberate character of changes in programme, in a word, all that places subjects within the system in a situation of unconscious effects of causality; all that produces effects of contingency between the action coming from the outside and the internal transformations of the system, all of that characterizes the non-genetic programme as much as the genetic one.”

Of course, Derrida is aware of this machinic force that haunts both writing and deconstruction and at times writes directly of it, offering it a proper name, but a proper name that has already been named as the opposing force of a radical other who would preclude the programmatic. In relation to Joyce’s Ulysses, referring to two Elijahs that inhabit the work, he states:
No longer Elijah the grand operator of the central, Elijah the head of the megaprogramotelephonic network, but the other Elijah, Elijah the other. But this is a homonym, Elijah can always be either one at the same time, one cannot call on one without risking getting the other.

In this sense deconstruction both requires this programmatic other and surrounds itself with a protective discourse that aims to resist the seduction of the programotelephonic Elijah that would make of deconstruction a machine that repeats the machine of metaphysics. One always risks the mutual contamination of these two Elijahs and in response deconstruction at places attempts to refuse the machinic Elijah. While Elijah the grand operator is never able to exclude or refuse Elijah the other, any who would ally themselves with the other must carefully avoid this force of alterity itself being haunted by a megaprogramotelephonic network. Derrida says: “I hear this vibration as the very music of Ulysses. The computer today cannot enumerate these interlacings, despite all the many ways it is already able to help us. Only a computer which has not yet been invented could answer that music in Ulysses.” For Derrida at the point of this text, no extant programming language or computer could ever integrate the beautiful ambiguity and musical vibrations of writing; only a computer to come.

Retaining the distinction between writing and program and with it a distinction between calculation and the incalcuable, between literature and the literal writing of programming has allowed a certain reading of deconstruction to conserve a relationship with authorial intent; as long as the program is neither deconstrucable nor a force of deconstruction, one can maintain, even if in covert fashion, that deconstruction must proceed exclusively as an authored theoretical activity. Minimizing the nonhuman agency of the program, and with it arche-writing, produces a corollary effect, reifying the sanctity of the subject who writes in human languages. This distinction is thus central to a deconstruction that would attempt to secure an authorial importance while disavowing the threat the sciences and maybe even life itself may portend to metaphysics.

Derrida, in speaking of animals, hints that the contamination of writing by animals and machines has always been central to his work: “This animal-machine has a family resemblance with the virus that obsesses, not to say invades everything I write. Neither animal nor nonanimal, organic or inorganic, living or dead, this potential invader is like a computer virus. It is lodged in a processor of writing, reading and interpretation.” Perhaps then all of deconstruction has been a question of this contamination between ‘natural’ languages and programs, of the machine and the human (and the animal somewhere between them).

Ultimately the program, especially as advances by cybernetics then transmutated into the advances of artificial intelligence, threatens in a way both to mark the ultimate victory of metaphysics, the pure presence of everything as data to computation, simulation and predication, but also its abandonment, the end of that adventure that associated technics with logos as Derrida says, its move beyond fixity and presence into a constantly shifting economy of différance, that is to say arche-writing and arche-violence as the non-ground of a post-metaphysical science, one that must first take up the work of deconstruction. Of course it is too early to even begin to sketch what such a science would look, I think if we are anywhere we are only at a moment where it may be possible to imagine whether or not such a science could be imagined.

To do so the language of writing, trace, grapheme, etc. must be thought as right on the edge of this deconstruction, a deconstruction which perhaps or perhaps not cybernetics threatened to traverse. As Derrida says, “The movements of belonging or not belonging to the epoch are too subtle, the illusions in that regard are too easy, for us to make a definitive judgement.” That is to say that it cannot be a choice simply between a deconstructive program or computational science or a metaphysical one, but rather the arrival and threat of both at once. I am reminded in this way of how Derrida ends structure, sign and play, contrasting a sad nostalgic Rousseauist interpretation with a joyful affirmative Nietzschean one but concluding that we are beyond the realm of being able to choose one or the other and rather are glimpsing the conception, formation and labor of a process of giving birth to the unameable and the non-species, “in the formless, mute, infant, and terrifying form of monstrosity.” If metaphysics as science, especially science of life, of the social, of the human and its others, is to give itself up it can only be in the form of the formless of the monstrous and unameable.

– Ann Arbor, February 20, 2020