“Computers of the world, unite!” is the inaugural essay of Persona non grata, a bilingual column by Jason Barker exclusively for Communis.
There’s a lot of shrill moralising going on about AI. Is it justified? Rather than attempt to answer that question let’s pose another: Will “AI” even be around in five years? If the means of communication have been migrated to the “clouds” then destiny surely awaits AI in the celestial sphere. Next stop Mars?
AI’s imperial designs on education are worth interrogating. Not for the dubious assertion that algorithms and data analysis can enhance the “learning experience”, but for what this gross intrusion into academia reveals about the precarious state of education today.
Recently a student from my department informed me that they had been accused of plagiarism. They protested their innocence to the professor, who responded by handing them an AI report stating that “we are moderately confident that this text is AI generated”. Whether or not the student was lying — they saw the funny side, which totally vindicates them in my book — seems less important than their sadly mistaken assumption that they were entitled to an education. In reality they had experienced something akin to that befuddled feeling you get when your credit card is declined at the grocery checkout for reasons unknown.
Professors might be forgiven for their cynicism. How many of us haven’t experienced that guilty pleasure when exposing the smart aleck at the back of the class? In any case, don’t we owe it to our students to uphold rigorous evidence-based standards of writing and research? Beyond question. What must be confronted, however, is the creeping prejudice on the part of non-academic administrators that academic rigour should yield to the “efficiency” of AI tools — somewhat like the domestic appliances that promised to “liberate” housewives in post-war America.
How to make sense of modernity’s obsession with formalization, with the universal balancing of inputs and outputs? Is there a formalist explanation for it? Or a technocratic one? Formalization is the axiomatic approach of dispensing with the peculiar properties of mathematical objects and boiling them down to sets of relations. Think systems. The formalist demands a self-consistent and all-encompassing set of rules for understanding how any given mathematical object will behave. Axiomatics is the proverbial Dr. Pangloss for whom humanity is an irreducibly corporate phenomenon incompatible with the “real experience” of any of its individuals. (Imagine a world in which economic activity is indifferent to human needs, where all value is equal and in perfect harmony, untroubled by the “epiphenomena” of the market, and you have a Marxist version of Pangloss).
Axiomatic thinking looms large in 20th century mathematics. David Hilbert’s announcement of “23 problems” at the International Congress of Mathematics held in Paris in 1900 — in which he invited mathematicians “to lift the veil behind which the future lies hidden; to cast a glance at the next advances of our science and at the secrets of its development during future centuries” — would launch an ambitious interdisciplinary program, led by Hilbert himself, which brought mathematics into dialogue with biology and theoretical physics, and which, despite being abandoned following Kurt Gödel’s groundbreaking work, has never ceased to generate scientific inventions.
Hilbert’s program would also pave the way for all manner of technocratic applications in the empirical sciences. Following a significant extension to axiomatic set theory and his major contributions to topology, algebra and statistics, John von Neumann departed the field of pure mathematics and proceeded to the axiomatization of quantum mechanics, before gravitating to hydrogen bomb design on the Manhattan Project. Some years later Alain Badiou would target such “interdisciplinarity” for its bogus science and ideological perversion of mathematical modeling. There is nothing scientific in devising a model for a world laboratory of mass murder. What concept of “world” is even at stake in the reproduction of empirical science, in its various iterations, beyond some misplaced idealism that we are living in the best of all possible worlds?
And computer science? Doesn’t it amount to yet another servile empiricism or idealism (both equivalent in Badiou’s eyes) for the modeling of a world of independent users, supposedly in direct and spontaneous communication with each other?
Computer science may be one of the most obfuscatory notions in scientific discourse. There is a pre-modern history of calculating machines and assorted technologies, from the abacus to the mass production arithmometer, which might be categorised as mechanical-arithmetic tools, or technologies of mental agility. The precursor to the abacus was a simple counting board (the Salamis tablet) which dates back to the 4th century BC, and in all likelihood provided a means of facilitating cross-cultural trade in the Mediterranean. (According to Parry Moon, the fact that counting boards were used by common merchants in Attica may explain why the aristocratic Plato avoids all mention of them.) By the 14th century the Japanese had developed the soroban, an abacus whose design included relatively large counting beads on short rods, presumably to allow for speedy calculation.
Arguably the modern history of computation begins with the concept of effective computability, which has nothing to do with the representational model—the ideology—of modern computing. The latter, like the merchant’s abacus, is geared towards achieving computational power and efficiency, and driven by the relentless abstraction of computational artefacts. Network computing via the internet has become synonymous with the expansion of knowledge (the “general intellect” or “hive mind”) which is completely unwarranted and somewhat ironic, given the prevailing tendency in software design towards the elimination (compression) of systemic redundancy. The idea of computational tools that facilitate the “user experience” has been surpassed by that of a virtual machine (i.e. software running relatively autonomously from its host and with minimal intervention) which renders the user a redundant concept, and any related “experience” a myth.
(A brief digression. In answer to Nick Bostrom’s question as to whether we’re living in a computer simulation: we just can’t tell, since the halting problem establishes that, for any given simulation (computer program), we cannot consistently decide whether it’s a simulation or not (i.e. that it halts) from within the simulation (i.e. for as long as the program is running). This is somewhat the hero’s dilemma in The Matrix. Given that Neo has spent his entire life living in a simulation, what rational basis does he have for believing that reality is any more real than the “simulation” he’s supposedly exited?)
Systemic redundancy is doubly ironic when applied to education. For over 50 years the humanities have been prophesying the “death of man” and the “death of the subject”, believing it was all a purely epistemological matter, a metaphor, something whose implications went no further than the seminar room. It’s as if professors, only now emerging from the wreckage of austerity and AI-assisted rationalisation, were saying to university management, or anyone likely to care, “Come on! You don’t think any of us were actually taking this stuff seriously, do you?” But people were, and the lack of any organisational strategy on the part of educators in the face of the AI wrecking ball simply mirrors (is a function of) their lack of any genuine concept of education.
Last year I set my students the task of summarising a chapter on Turing machines. Roughly fifty percent of them fed the chapter into an AI summarizer, a fact confirmed to me by AI detection software. What the AI cohort really should have done, I informed the class, in order to avoid detection, was to use an AI detection avoidance summarizer—although, as I went on to explain, that wouldn’t necessarily help them, since I could just as easily feed their summaries into an AI detection avoidance summarizer detector (a defining feature of the universal Turing machine is its ability to recursively simulate the behaviour of any other Turing machine).
It’s certainly “ironic” that students, when asked “what is a Turing machine?”, would opt to ask a Turing machine. If the Turing machine provides the answer, not the student, then obviously the student has learned nothing. And yet perhaps the more worrying development here is that neither has the Turing machine. “AI” (I hesitate to cast aspersions on artificial intelligence) cannot know that it knows, despite the fact that it can “know” (imitate) its human interlocutors, as Alan Turing-the-man set out in his famous paper on “Computing Machinery and Intelligence”, where he devises the so-called Turing test.
Now, whether the fact of Turing machines and the Turing test can provide succor to educators facing AI disruption in their classrooms, I wouldn’t like to say. Could we envision a future for the humanities in which textual analysis becomes a sort of cryptography? Or where “creative writing” involves using AI to brainstorm experimental prose (that in six months’ time AI will be writing itself, thus making the concept of creative writing redundant)? Certainly. At least for as long as humanities departments can generate sufficient new courses to keep the fires of AI creativity burning. But a word of caution: AI does creative destruction much better than we do, free of all the moral sermonising, and without the need to justify to its stakeholders the “relevance” of its courses.
Moving away from this technocratic model of computer science, with its systemic redundancy and built-in obsolescence (update your OS, divulge your personal data, renounce your right to privacy, or else be locked out of your own life) let’s return to the following point, namely, that the modern history of computation begins with the concept of effective computability, a term Alan Turing coined in response to Hilbert and Ackermann’s Entscheidungsproblem (decision problem) of 1928 (Hilbert and Ackermann’s problem had in turn been inspired by Leibniz’s dream of inventing a universal calculator).
In more straightforward terms, “computing machine” is Turing’s description of a hypothetical model for the computation of any computable number. It’s not an actual physical machine requiring an integrated operating system for managing and prioritising tasks. And yet the “simplicity” of Turing’s computing machine outweighs by far the “complexity” of AI, since the former is solely concerned with computing the limits of computation. Indeed, in terms of classes of automata, with combinational logic at the low end and Turing machines at the high end, AI isn’t even on a par with pushdown automata, since AI algorithms are largely confined to pattern recognition and basic decision-making, and require far more training to achieve what pushdown automata achieve by definition (namely, recognising languages with nested structures).
There was a time, in the halcyon days of the civil service, when “the computer” was a human being, invariably equipped with basic measuring devices (a microscope, stopwatch, tally counter) but perhaps most often with nothing more technical than paper and pencil. Inevitably most computers were women and poorly paid. Did this fact influence Turing to come up with what is essentially a proletarian model of computation, based on the predicament of the unskilled and alienated worker of the post-war era? If we consider that what Turing sets about proving in his Entscheidungsproblem response paper of 1936 is that anyone with access to a set of instruction tables—and providing they can read them—will be capable of carrying out any mathematical calculation, then what he comes up with is truly an egalitarian manifesto-in-the-making for unskilled intellectual workers.
And time, in the halcyon days of the civil service, was at least somewhat on the computer’s side. Effective computability is by no means determined by computational efficiency. If anything, it’s the reverse, with the computer computing until such time as the task is “finished”, solely on the basis of the instructions required for the precise completion of the task. The work takes as long as it takes, from each according to their ability, to each according to their needs.
Imagine a world where the computer was not so much equipped with as defined by the social and political potentials of computation. What sort of world might that be?
Jason Barker is one of the most outstanding prose stylists of the past 15 minutes.
Featured image: Electronic Numerical Integrator and Computer (ENIAC), the world’s first computer. Photo courtesy the U.S. Army and the archives of the ARL Technical Library for SIMS. LIFECYCLE SERVICES.
