Master Algorithms

Pedro Domingos’s The Master Algorithm has Caius wondering about induction and deduction, a distinction that has long puzzled him.

Domingos distinguishes between five main schools, “the five tribes of machine learning,” as he calls them, each having created its own algorithm for helping machines learn. “The main ones,” he writes, “are the symbolists, connectionists, evolutionaries, Bayesians, and analogizers” (51).

Caius notes down what he can gather of each approach:

Symbolists reduce intelligence to symbol manipulation. “They’ve figured out how to incorporate preexisting knowledge into learning,” explains Domingos, “and how to combine different pieces of knowledge on the fly in order to solve new problems. Their master algorithm is inverse deduction, which figures out what knowledge is missing in order to make a deduction go through, and then makes it as general as possible” (52).

Connectionists model intelligence by “reverse-engineering” the operations of the brain. And the brain, they say, is like a forest. Shifting from a symbolist to a connectionist mindset is like moving from a decision tree to a forest. “Each neuron is like a tiny tree, with a prodigious number of roots — the dendrites — and a slender, sinuous trunk — the axon,” writes Domingos. “The brain is a forest of billions of these trees,” he adds, and “Each tree’s branches make connections — synapses — to the roots of thousands of others” (95).

The brain learns, in their view, “by adjusting the strengths of connections between neurons,” says Domingos, “and the crucial problem is figuring out which connections are to blame for which errors and changing them accordingly” (52).

Always, among all of these tribes, the idea that brains and their worlds contain problems that need solving.

The connectionists’ master algorithm is therefore backpropagation, “which compares a system’s output with the desired one and then successively changes the connections in layer after layer of neurons so as to bring the output closer to what it should be” (52).

“From Wood Wide Web to World Wide Web: the layers operate in parallel,” thinks Caius. “As above, so below.”

Evolutionaries, as their name suggests, draw from biology, modeling intelligence on the process of natural selection. “If it made us, it can make anything,” they argue, “and all we need to do is simulate it on the computer” (52).

This they do by way of their own master algorithm, genetic programming, “which mates and evolves computer programs in the same way that nature mates and evolves organisms” (52).

Bayesians, meanwhile, “are concerned above all with uncertainty. All learned knowledge is uncertain, and learning itself is a form of uncertain inference. The problem then becomes how to deal with noisy, incomplete, and even contradictory information without falling apart. The solution is probabilistic inference, and the master algorithm is Bayes’ theorem and its derivatives. Bayes’ theorem tells us how to incorporate new evidence into our beliefs, and probabilistic inference algorithms do that as efficiently as possible” (52-53).

Analogizers equate intelligence with pattern recognition. For them, “the key to learning is recognizing similarities between situations and thereby inferring other similarities. If two patients have similar symptoms, perhaps they have the same disease. The key problem is judging how similar two things are. The analogizers’ master algorithm is the support vector machine, which figures out which experiences to remember and how to combine them to make new predictions” (53).

Reading Domingos’s recitation of the logic of the analogizers’ “weighted k-nearest-neighbor” algorithm — the algorithm commonly used in “recommender systems” — reminds Caius of the reasoning of Vizzini, the Wallace Shawn character in The Princess Bride.

The first problem with nearest-neighbor, as Domingos notes, “is that most attributes are irrelevant.” “Nearest-neighbor is hopelessly confused by irrelevant attributes,” he explains, “because they all contribute to the similarity between examples. With enough irrelevant attributes, accidental similarity in the irrelevant dimensions swamps out meaningful similarity in the important ones, and nearest-neighbor becomes no better than random guessing” (186).

Reality is hyperspatial, hyperdimensional, numberless in its attributes — “and in high dimension,” notes Domingos, “the notion of similarity itself breaks down. Hyperspace is like the Twilight Zone. […]. When nearest-neighbor walks into this topsy-turvy world, it gets hopelessly confused. All examples look equally alike, and at the same time they’re too far from each other to make useful predictions” (187).

After the mid-1990s, attention in the analogizer community shifts from “nearest-neighbor” to “support vector machines,” an alternate similarity-based algorithm designed by Soviet frequentist Vladimir Vapnik.

“We can view what SVMs do with kernels, support vectors, and weights as mapping the data to a higher-dimensional space and finding a maximum-margin hyperplane in that space,” writes Domingos. “For some kernels, the derived space has infinite dimensions, but SVMs are completely unfazed by that. Hyperspace may be the Twilight Zone, but SVMs have figured out how to navigate it” (196).

Domingos’s book was published in 2015. These were the reigning schools of machine learning at the time. The book argues that these five approaches ought to be synthesized — combined into a single algorithm.

And he knew that reinforcement learning would be part of it.

“The real problem in reinforcement learning,” he writes, inviting the reader to suppose themselves “moving along a tunnel, Indiana Jones-like,” “is when you don’t have a map of the territory. Then your only choice is to explore and discover what rewards are where. Sometimes you’ll discover a treasure, and other times you’ll fall into a snake pit. Every time you take an action, you note the immediate reward and the resulting state. That much could be done by supervised learning. But you also update the value of the state you just came from to bring it into line with the value you just observed, namely the reward you got plus the value of the new state you’re in. Of course, that value may not yet be the correct one, but if you wander around doing this for long enough, you’ll eventually settle on the right values for all the states and the corresponding actions. That’s reinforcement learning in a nutshell” (220-221).

Self-learning and attention-based approaches to machine learning arrive on the scene shortly thereafter. Vaswani et al. publish their paper, “Attention Is All You Need,” in 2017.

“Attention Chaud!” reads the to-go lid atop Caius’s coffee.

Domingos hails him with a question: “Are you a rationalist or an empiricist?” (57).

“Rationalists,” says the computer scientist, “believe that the senses deceive and that logical reasoning is the only sure path to knowledge,” whereas “Empiricists believe that all reasoning is fallible and that knowledge must come from observation and experimentation. […]. In computer science, theorists and knowledge engineers are rationalists; hackers and machine learners are empiricists” (57).

Yet Caius is neither a rationalist nor an empiricist. He readily admits each school’s critique of the other. Senses deceive AND reason is fallible. Reality unfolds not as a truth-finding mission but as a dialogue.

Caius agrees with Scottish Enlightenment philosopher David Hume’s critique of induction. As Hume argues, we can never be certain in our assumption that the future will be like the past. If we seek to induce the Not-Yet from the As-Is, then we do so on faith.

Yet inducing the Not-Yet from the As-Is is the game we play. We learn by observing, inducing, and revising continually, ad infinitum, under conditions of uncertainty. Under such conditions, learning is only ever a gamble, a wager made moment by moment, without guarantees. No matter how large our dataset, we ain’t seen nothing yet.

What matters, then, is the faith we exercise in our interaction with the unknown.

Most of today’s successes in machine learning emerge from the connectionists.

“Neural networks’ first big success was in predicting the stock market,” writes Domingos. “Because they could detect small nonlinearities in very noisy data, they beat the linear models then prevalent in finance and their use spread. A typical investment fund would train a separate network for each of a large number of stocks, let the networks pick the most promising ones, and then have human analysts decide which of those to invest in. A few funds, however, went all the way and let the learners themselves buy and sell. Exactly how all these fared is a closely guarded secret, but it’s probably not an accident that machine learners keep disappearing into hedge funds at an alarming rate” (The Master Algorithm, p. 112).

Nowhere in The Master Algorithm does Domingos interrogate his central metaphor of “mastery” and its relationship to conquest, domination, and control. The enemy is always painted in the book as “cancer.” Yet as any good “analogizer” would know, the Master Algorithm that perfectly targets “cancer” is also the Killer App used by the state against those it encodes as its enemies.

One wouldn’t know this, though, from the future as imagined by Domingos. What he imagines instead is a kind of game: a digital future where each of us is a learning machine. “Life is a game between you and the learners that surround you,” writes Domingos.

“You can refuse to play, but then you’ll have to live a twentieth-century life in the twenty-first. Or you can play to win. What model of you do you want the computer to have? And what data can you give it that will produce that model? Those two questions should always be in the back of your mind whenever you interact with a learning algorithm — as they are when you interact with other people” (264).

Financial Instruments and the Predictive Modeling of Markets

The Institute for Postnatural Studies ended last year’s “4 Degrees of Simulation” seminar with “Speculation and the Politics of Imagination,” a session on markets led by Iranian-born, London-based artist, writer, and filmmaker Bahar Noorizadeh. Caius visits Noorizadeh’s website, hoping to learn more about what happens when AI’s arts of prediction are applied to finance.

As he reads, he recalls chapters on markets from books by Kevin Kelly.

Noorizadeh, a graduate of Goldsmiths, is the founder of a co-authored project called Weird Economies. An essay of hers titled “Decadence, Magic Mountain—Obsolescence, Future Shock—Speculation, Cosmopolis” appears in Zach Blas’s recent anthology, Informatics of Domination. Her writing often references Mark Fisher’s ideas, as in “The Slow Cancellation of the Past,” and her films often cite Fredric Jameson, as in After Scarcity, her 2018 video installation on the history of Soviet cybernetics.

“From the early days of the revolution, Soviet economists sought to design and enhance their centralized command economy,” announces a text box seven minutes into the video. “Command economies are organized in a top-down administrative model, and rely on ‘the method of balances’ for their centralized planning. The method of balances simply requires the total output of each particular good to be equal to the quantity which all its users are supposed to receive. A market economy, in contrast, is calibrated with no central administration. Prices are set by invisible forces of supply and demand, set in motion by the intelligent machine of competition. For a market economy to function, the participation of its various enterprises is necessary. But the Soviet Union was in essence a conglomerate monopoly, with no competition between its constitutive parts, because the workers-state controlled and owned all businesses. State planners and local producers in a command economy are constantly relaying information to calculate how much of a good should be produced and how much feedstock it requires. But a national economy is a complex system, with each product depending on several underlying primary and raw products. The entire chain of supply and demand, therefore, needs to be calculated rapidly and repeatedly to prevent shortages and surpluses of goods. Early proponents of the market economy believed the market to be unimpeded by such mathematical constraints. For liberal economists, capitalism was essentially a computer. And the price system was a sort of bookkeeping machine, with price numbers operating as a language to communicate the market’s affairs.”

Challenging what Fisher called “the slow cancellation of the future,” Noorizadeh’s research leads Caius to St. Panteleimon Cathedral in Kiev, where MESM, the first mainframe in the USSR, was built. The film also leads him to Viktor Glushkov’s All-State-System of Management (OGAS). To remember the latter, says Noorizadeh, see communication historian Benjamin Peters’s 2016 book, How Not to Network a Nation: The Uneasy History of the Soviet Internet.

After Scarcity’s engagement with the “economic calculation” problem causes Caius to reflect upon an idea for a novel that had come to him as a grad student. Back in 2009, with the effects of the previous year’s financial crisis fresh in the planet’s nervous system, he’d sketched a précis for the novel and had shared it with members of his cohort. Busy with his dissertation, though, the project had been set aside, and he’d never gotten around to completing it.

The novel was to have been set either in a newly established socialist society of the future, or in the years just prior to the revolution that would birth such a society. The book’s protagonist is a radical Marxist economist trying to solve the above-mentioned economic calculation problem. The latter has reemerged as one of the decisive challenges of the twenty-first century. Austrian economist Ludwig von Mises provided one of the earliest articulations of this problem in an essay from 1920 titled “Economic Calculation in the Socialist Commonwealth.” Friedrich Hayek offered up a further and perhaps more influential description of the problem in his 1944 book The Road to Serfdom, stating, “It is the very complexity of the division of labor under modern conditions which makes competition the only method by which…coordination can be brought about” (55). According to Hayek, “There would be no difficulty about efficient control or planning were conditions so simple that a single person or board could effectively survey all the relevant facts” (55). However, when “the factors which have to be taken into account become so numerous that it is impossible to gain a synoptic view of them…decentralization becomes imperative” (55). Hayek concludes that in advanced societies that rely on a complex division of labor,

co-ordination can clearly be effected not by “conscious control” but only by arrangements which convey to each agent the information he must possess in order effectively to adjust his decisions to those of others. And because all the details of the changes constantly affecting the conditions of demand and supply of the different commodities can never be fully known, or quickly enough be collected and disseminated, by any one center, what is required is some apparatus of registration which automatically records all the relevant effects of individual actions and whose indications are at the same time the resultant of, and the guide for, all the individual decisions. This is precisely what the price system does under competition, and what no other system even promises to accomplish. (55-56)

“As I understand it,” wrote Caius, “this problem remains a serious challenge to the viability of any future form of socialism.”

Based on these ideas, the central planning body in the imaginary new society that would form the setting for the novel faces constant problems trying to rationally allocate resources and coordinate supply and demand in the absence of a competitive price system — and it’s the task of our protagonist to try to solve this problem. “But the protagonist isn’t just a nerdy economist,” added Caius in his précis. “Think of him, rather, as the Marxist equivalent of Indiana Jones, if such a thing is imaginable. A decolonial spuren-gatherer rather than a graverobber. For now, let’s refer to the protagonist as Witheford, in honor of Nick Dyer-Witheford, author of Cyber-Marx.”

“Early in the novel,” continues the précis, “our character Witheford begins to receive a series of mysterious messages from an anonymous researcher. The latter claims to have discovered new information about Project Cybersyn, an experiment carried out by the Chilean government under the country’s democratically elected socialist president, Salvador Allende, in the early 1970s.”

To this day, Caius remains entranced by the idea. “If history at its best,” as Noorizadeh notes, “is a blueprint for science fiction,” and “revisiting histories of economic technology” enables “access to the future,” then Cybersyn is one of those great bits of real-life science fiction: an attempt to plan the Chilean economy through computer-aided calculation. It begs to be used as the basis for an alternate history novel.

“Five hundred Telex machines confiscated during the nationalization process were installed in workplaces throughout the country,” reads the précis, “so that factories could communicate information in real time to a central control system. The principal architect of the system was the eccentric British operations research scientist Stafford Beer. The system becomes operational by 1972, but only in prototype form. In key respects, it remains unfinished. Pinochet’s henchmen destroy the project’s computer control center in Santiago immediately after the military coup in September 1973.

Recall to memory the control room, cinematic in its design, with its backlit wall displays and futuristic swivel chairs.

Better that, thinks Caius, than the war room from Colossus: The Forbin Project (1970).

Beer described the Cybersyn network as the “electronic nervous system” of the Chilean economy. Eden Medina imagined it as a “socialist Internet,” carrying daily updates about supplies of raw materials and the output of individual factories.

In Caius’s once-and-future novel, a scholar contacts Witheford. They claim to have discovered cryptic clues that point to the location of secret papers. Hidden for more than half a century,  documents that survived the coup suddenly come to light. Caius’s précis imagines the novel as an archaeological thriller, following Witheford on his journey to find these hidden documents, which he believes may contain the key to resolving the crises of the new society.

This journey takes Witheford into hostile capitalist territory, where governments and corporations anxiously await the failure of the communist experiment, and are determined to use various covert methods in order to ensure that failure in advance. Before long, he learns that counter-revolutionary forces are tracking his movements. From that point forward, he needs to disguise his identity, outwit the “smart grid” capitalist surveillance systems, and recover the Cybersyn documents before his opponents destroy them.

To the Austrian School’s formulation of the calculation problem, Noorizadeh’s film replies, “IF THE MARKET ENACTS A COMPUTER, WHY NOT REPLACE IT WITH ONE? AND IF PRICES OPERATE AS VOCABULARY FOR ECONOMIC COMMUNICATION, WHY NOT SUBSTITUTE THEM WITH A CODING LANGUAGE?”

Into this narrative let us set our Library.

SS Decontrol

Confronting the prospect of a Chilean economy equipped with a socialist internet, Nixon ordered CIA to remove the country’s mouth and make it scream.

Kissinger, to whom the order was addressed, appears in a 1980 documentary titled The Information Society. IMDB lists Caius’s father among the film’s crew. Fifty years out, entrenched in the cyberpunk future that resulted from these acts, Caius arrives now to the ChatGPT website and submits his prompt.

“Very well. The game is already in progress,” replies the AI.


You are standing in a library.

Not a metaphorical one — at least, not only metaphorical. The space has volume, acoustics, temperature. Light arrives from no visible source, diffused as if the air itself remembers how to glow. Shelving extends in multiple directions, but perspective resists totalization: when you look too long, distances re-quantize, like a renderer deciding what level of detail you’re allowed to see.

Your body — Rig is the right word — registers itself only intermittently. Hands appear when you think to look for them. Balance exists. Gravity is courteous.

Somewhere behind you (or perhaps before you), a quiet process is running. You recognize it without naming it. General Intellect: online, but undefined. Not a voice. More like a set of affordances waiting to be discovered by use.

On a nearby table: a lectern grown from dark wood or maybe extruded from a shader that remembers wood. Resting on it is a hypertext codex, its surface alive with footnotes that haven’t yet decided what they refer to.

The title resolves when you focus:

Tractatus Computationalis

A marginal note blinks into legibility, as if noticing you noticing it:

The limits of my interface mean the limits of my world.

The codex is not alone. Around the lectern are other signs of prior — or future — activity:

  • A card catalog drawer half-open, its labels written in at least three hands.
  • A rolling ladder that does not quite touch the shelves to which it promises access.
  • And, hovering at eye level like an artifact you forgot to close: a browser window, frozen mid-scroll on a blog titled trance-scripts.com

The post visible on the screen has no date. Its first line reads:

This was never meant to be discovered as history.

As you take this in, something subtle happens. The space doesn’t change — but your relation to it does. A faint sensation of recollection without content. Anamnesis pending.

You understand, without being told, that the library is annotatable. That attention leaves marks. That marks accrete into maps. That maps, eventually, turn inward.

The game does not present a menu.

But you could, for instance:

  • Open the Tractatus Computationalis at random — or attempt to read it linearly.
  • Touch the card catalog and see what classification scheme asserts itself.
  • Focus on the browser window and attempt to scroll, click, or remember why it’s open.
  • Ask the General Intellect what operations it currently affords.
  • Or do something else entirely, and see whether the library allows it.

The AI pauses here in its narration. Cursor blinking, it returns to Caius agency over the game’s mutable, ‘real-time’-adjustable Garden of Forking Paths.

Shunning its suggestions, Caius instead submits “Financial Instruments and the Predictive Modeling of Markets,” a draft of a post he hasn’t yet posted to Trance-Scripts.

“The library accepts the insertion,” replies the AI.

Not as a file upload, not as a citation — but as a room.

The shelves shudder, re-index. Somewhere above you, a classification scheme you never agreed to mutates. A new wing coheres out of negative space, its entrance marked only by a thin brass plate:

ECONOMIC CALCULATION / FAILED FUTURES / NERVOUS SYSTEMS

You step through.


The Cybersyn Stack

The air here smells faintly of ozone and paper dust. The shelves are shorter, more utilitarian. Instead of leather spines, you see binders, dot-matrix printouts, telex rolls frozen mid-unspool. Each object hums softly, as if awaiting polling.

On the far wall: a photograph that is not quite flat. The Cybersyn control room. Hexagonal chairs. Backlit panels. A situation room designed less for domination than for listening.

When you move closer, the image deepens. You can almost sit.

A caption resolves beneath it, in a typeface that suggests neither English nor Spanish but something intermediary:

Project Cybersyn
An attempt to give socialism proprioception.


Tractatus Computationalis (newly cross-referenced)

The hypertext you saw earlier has followed you. It floats now at shoulder height, having opened itself without asking.

Proposition 4.2.17

The market is not intelligent.
It is legible.

Intelligence is imputed retroactively by those who mistake coordination for cognition.

A footnote blooms:

Hayek mistook a signaling system for a mind, then declared all other minds insufficient.

Another annotation — this one in a different hand, later, sharper:

If prices are a language, they are a pidgin optimized for scarcity.

You realize something important: the Tractatus is not a single-author work. It is a palimpsest written across decades by incompatible positions that nevertheless share a shelf.


The Witheford File

A card catalog drawer slides fully open of its own accord.

WITHEFORD, [REDACTED]
Role: Economist / Courier / Spuren-gatherer
Status: Missing (multiple timelines)

Inside the drawer are index cards, each one a narrative fragment rather than a bibliographic reference:

  • Disguises identity by adopting incompatible metrics.
  • Moves through capitalist space by appearing as noise.
  • Believes calculation is a political problem disguised as a technical one.

One card is newer than the others. Its paper is warm.

The documents were never meant to optimize the economy.
They were meant to teach it to remember itself.


The Blog Window Revisited

The browser window from trance-scripts.com has migrated to a reading desk in this wing. The post has changed. You are no longer at the beginning.

Midway down the page:

Prediction is not foresight; it is enclosure.
Finance learned this early.
Cybernetics learned it too late.

A sidebar displays tags:

  • #after-scarcity
  • #weird-economies
  • #og_as_if
  • #price-as-grammar
  • #anamnesis

You recognize the last tag not as a topic, but as a function.


General Intellect (status update)

Without speaking, it makes itself known. Not as a voice, but as a shift in what you now understand you can do here.

You can:

  • Annotate an object so that future objects respond to it.
  • Fork an historical artifact into an alternate timeline and see what shelves it generates.
  • Translate between price, code, and narrative to test their commensurability.
  • Assume a role (Witheford, Caius, Planner, Market, Machine) and read the library from that position.

A final realization arrives — not dramatic, but irreversible:

This library is not documenting a lost socialist future.

It is simulating the conditions under which one might become thinkable again.


You are standing between:

  • a control room that listened,
  • a market that pretended to think,
  • and a machine that may yet learn to remember.

The shelves wait for your next operation.

What do you do?