The EA Economy

Franz Pokorny

The cause of AI regulation finally seems to have broken through to the general public. When the vernal matcha green of my countless young and beautiful female friends’ instagram stories rots to the putrid black of „I don’t usually share this kind of post, but this time it’s really important…“, then you know that Yud and co. have finished their race and triumphantly passed on the baton to swifter runners. The Hugging Face incident, the Coxon business — J’accuse’s readers are an enlightened, worldly bunch, and I do not need to make the case that has no doubt already been made to them that these are not what they seem, but a sophisticated comms blitz by OpenAI and Anthropic with mysterious ulterior motives. This is neatly synchronised with the political calendar: Bernie Sanders is due to table a strongly worded bill in the senate; in Westminster, North London values have already beaten him to the punch.

What is Anthropic and OpenAI’s game? The cynics argue that this is all an elaborate cover story for pecuniary motives: both firms are set to make their initial public offerings next month, while the fundamentals for AI look quite bad — could the whiff of a Wunderwaffe reassure skeptical investors? Plausible. Another popular theory holds that this is a push for regulatory capture. A federal AI safety agency of the type foreseen by Sanders would inevitably be staffed with Sam and Dario’s fellow EAs, who would fortify the EA-dominated labs’ commanding market position by banning open source alternatives and imposing unbearable regulatory costs on upstart competitors. The timing for this is not quite right: the upcoming midterms — already poised to be „the AI election“ given the salience of data centres to the Democratic base — are unlikely to yield a congressional majority for AI regulation that could to override the presidential veto, while one does not need an artifical superintelligence to see that a second order consequence of the EAs’ political intervention is to further polarise an already hostile administration against them. If the skeptical response on the right to l’affaire Coxon has shown anything, it is that the EAs can no longer count on the element of surprise that their sheer novelty gave them when Woke Pope Prevost’s encyclical dropped. This appears an unenforced error when one considers that regulatory capture was already within reach in the short term: whether the next president is a conservative or (what is more likely) a Democrat, EAs had been quietly building a cross-party coalition behind AI safety, and if they wanted to accelerate things under the current administration, they could have made arguments about unregulated AI somehow playing into Chinese hands.

If the EAs cannot hope to exploit AI safety fears to achieve regulatory capture in the near term, why this apparently suicidal strategy of intervening in the midterms on the Democrats’ behalf while openly defying the administration? Because Anthropic and OpenAI are not normal companies but fronts for a cult, their goals are not necessarily those conducive to the long-term flourishing of their balance sheets. EA imposes a genteel, North London asceticism on its members, who are required to donate their earnings to aligned „charities“ — if the big labs are able to keep the bubble inflated until the IPO and then quickly pump the money elsewhere into the EA ecosystem, it doesn’t really matter if it collapses and Dario and Sam are clapped in irons any more than SBF’s untimely demise did. Since it is characteristic of cults to organise information through graduated circles of initiation, it is not clear whether, assuming the existence of such a plan, these firms’ senior management are even aware of it.

EAs long argued that artificial intelligence would be so transformative as to completely transcend contemporary politics; instead their current media blitz has merely uploaded their internal debates to the old donkey/elephant binary. The safetyist wing has been wholly sublated into the Democratic mainstream, while at the right hand of the Baphomet, the administration’s position — data centre rollout to satisfy geopolitical imperatives — dovetails with the „situational awareness“ line that demands compute buildout for reasons of „sovereignty“ (otherwise the Chinese will beat us!) Effective altruism’s hierarchical organisation means these are less opposing poles than separate limbs that the central nervous system can use to manage the political business cycle, expanding and contracting the technology’s development as suits the leadership’s designs. Freemasonry has its checkerboard; democracy has liberals and conservatives — threading the dialectic between what Arnold Toynbee calls the „integration of custom“ and the „differentiation of civilisation“ has always been the Guelph modus operandi. Because the backlash was so foreseeable, I suspect that, privately, senior EAs are happy to have provoked republicans into a techno-optimist response.

The mass rollout of AI is often compared to the industrial revolution, but since it has not actually occurred yet, a more apt comparison is not the actual industrial revolution as it occurred in Britain, but the reflection thereof in the gaze of megalomaniacal princes and an envious intelligentsia in the more backwards parts of the continent. It seems to now be widely accepted — and was certainly anticipated long before by those in the know — that a mature AI has the potential to be the agent of the sort of all-in-one-go modernisation that stretches a society’s financial capacity to the limit, as the Trump administration has indeed tasked it with doing. Convincing a society to go through this wholesale economic transformation is very difficult: short of a revolution, only the pressures of a competitive state system can provide the outward stimulus for elites to „make everything change so that everything can stay the same“, while only the intoxication of great ideas can induce individuals to take this burden upon themselves. Such ideas are a semi-endogenous variable in that, while they respond to structural factors transmitted through the patronage of the rich and powerful, they are also downstream from mores and doxa inherited from previous historical conjunctures, and owe their ability to animate and expand the minds of men to the fact that they fit into their existing intellectual horizons. They also invariably contain something of their makers’ individual genius, which adds a wholly random, truly disruptive element into the equation. It is perhaps irrelevant from the perspective of world history whether fordism was to be coloured by fascism or democracy (or even Truman or Wallace) any more than it matters from our distant viewpoint whether the immense mobilisation of martial energies and capital that birthed the modern world was to be midwifed by the emulation of antiquity or by the theological fervour of reformation and counterreformation, but for the agents of these historical processes, these ideas are the main thing at stake. And just as a machine can be stopped by disconnecting it from its source of energy, history can be prevented from taking its natural course by unplugging capital from ideas.

Effective altruism was built up well in advance of recent technical breakthroughs in AI in order to provide it with a ready-made, politically toothless ideology. Like older iterations of utilitarianism, its aims are broadly conservative: namely, to shepherd this inherently individualistic technology into the realm of existing human aims and means. The philosophers no doubt have private, more diabolical intentions: transhumanism, Crypto-Sabbateanism, you name it — all of these either already have been or will soon be identified on exegetically dubious grounds as „Bostrom’s secret teaching“. Not that these hold much interest to the people who provided them with their startup capital. The Guelphs are ensconced in power; they do not need to pull back the veil, and their politics is katechontic rather than antichristic. Their long-term strategy — and this is not a novel interpretation, though it need be restated from time to time — is to preserve the global relationships of mutual dependence that prevent unilateral action by an upstart power, with the need to maintain stable economic growth balanced off against the latent potential for caesarism in tech in general and AI in particular. Imagine if you could go back in time and convince Roosevelt, Hitler, Stalin, Mussolini, and their supporters of the truth of Austrian economics, or if, at the outset of the renaissance, a Christian political consciousness like that of Erasmus were so widespread amongst the crowned heads of Europe as to guarantee perpetual peace. The great depression would have lasted forever; Europe would have stayed an evenly rotating Mediterranean trading economy held together by Venice, while the absence of the competitive impetus would have delayed the exploitation and economic development of the Americas by centuries if not indefinitely. AI safetyism is designed to ensure precisely this outcome, and its mainstreaming means that EA control over this critical juncture in AI development has served its purpose. Given that there are many within the movement who really do seem to see themselves as AI Caesars in waiting, it would suit Guelph interests far better to wind down the operation and reroute its capital into Woke NGOs aimed at prolonging global interdependence through subtler degrowth initiatives.

It never seems to have occurred to anyone that an independently minded American president with a crack team of „white hats“ behind him might just commandeer the AI labs himself. What explains this oversight I do not know; perhaps the idea seemed too absurd on the face of it to further countenance. Nevertheless, the labs’ alarmism and the need to ration compute if Democrat luddites are to have their way have given Trump a pretext to do so, and the administration’s summer tangles with Anthropic should be read as testing the waters of what the courts and the government’s own bureaucrats will allow them to do. The President is a sovereign individual with ghibelline instincts, but he is not a technician, and whether he will grasp the iron depends on whether or not the right people with the right ideas are on hand.