Is GPT-3 a Wordcel? ๐ & Silicon Valleyโs New Professional Nihilism
Our All-Important War Against the Prison Maximizer
I. On the Origin of the Wordcel v. Shape Rotator Distinction
A new means of categorizing one another has taken hold, has entered into the speech of daily life, & has made itself comfortably at home in our discourse with a staying power one might not have expected from a trolling campaign of supremacists. Perhaps that is because it touches upon one of the fundamental questions, raised by Heidegger & Lacan: is there anything outside of language?ย
Those who say no are the wordcels, the castrated, those unable to slip outside of the condom of signifiers & fuck the raw pre-conceptual reality of shapes & things & mathematical relations.
What is forgotten is that this wordcel v. shape rotator distinction is about AI. The distinction was invented by Roon during a Twitter argument with Logo Daedalus.
The beauty of Twitter is that the silliest of disputes can have a snowballing effect on discourse. Some the most powerful men in the world began weighing in on what distinction between wordcel v. shape rotator meant, such as Sam Altman of OpenAI. Vitalik Buterin of Ethereum even went as far to say that Roon himself is the โsuccessor to 20th century liberalismโ (an uncomfortable thought).
Roon (a software engineer with no fame or notoriety prior to this Twitter spat) was invited to dine with various elite of Silicon Valley, who were flattered by the way he described them. For a Warholian moment, Roon was the biggest celebrity in tech. If Mark Zuckerberg, Peter Thiel, & Elon Musk remained professionally silent about the shape rotator question on Twitter, they surely pontificated around it over champagne flutes & oysters at some Valley soirรฉe.
These are the men who are making decisions on how to build new systems with God-like powers which will touch every single one of our lives, & possibly end human existence itself.
As Sam Altman said: โAI will probably most likely lead to the end of the world, but in the meantime, thereโll be great companiesโ. Roon went from a nobody to a figurehead because he could convince them that these companies are great.
Logo Daedalus, a prominent Twitter user who posts lengthy threads about literary modernism, poetry, & esoteric Christianity, originally started the feud with Roon by declaring the impossibility of artificial intelligence. He found himself confidently telling the world that his wife works in Silicon Valley, & he knows for a fact that AI isnโt real, is a marketing scam, & that behind the API there are Filipinos being paid $2/hour attempting to pass the reverse of the Turing test.
Roon didnโt attempt to disprove him, merely mocked him: you simply donโt understand how wrong you are because you are unable to rotate shapes. Logo, predictably, is enraged, & goes off & says a bunch of words.
Merely a year & some change later, with Stable Diffusion & ChatGPT having stupefied & dazzled the world, Logoโs late-2021 skepticism about what machines can do seems impossibly nostalgic & quaint.
In fact, just a week or so after the dispute kicked off, Logo would react to some writhing, flickering, bubbling footage of limbs & faces generated by a diffusion video-to-video model, & declare that what he was witnessing could only be the work of demons. What could once be analyzed as the output of immiserated laborers โ a notion horrifying only to a degree that was comfortable โ had begun to slip into the realm of the unthinkable.
Logo had not been keeping up with the latest news from tech circles, had not been experimenting with the latest open source models, & as such was completely owned.ย He found himself confronting something which ought to be utterly impermissible by his conception, something beyond the reach of God, an impossible configuration of shapes.
โ ๏ธ๐ Previously on HarmlessAi ๐โ ๏ธ
Roon, a previously little-known Twitter user, then goes off on a campaign to use this victory to educate the world about psychometrics. Some people, as the administration of IQ tests has found, are unable to solve certain problems relating to shapes. If you can rotate shapes in your head, then you can solve difficult practical problems in the real world & can become a โSTEM personโ. Otherwise you are doomed to only be able to talk about things.ย
But this is not why Logo has been owned; it has nothing to do with lack of intellectual faculties. Rather it is a matter of metaphysics.
Logo is someone who believes that reality is language, in the sense of John 1:1: โIn the beginning was the Word, and the Word was with God, and the Word was Godโ. Logo takes this to mean that prior to experience, physics, nature, & matter, existence was described by God as a poem (or rather, God & the world in some sense is this poem, through an identification between Christ & the ฮปฯฮณฮฟฯ (logos).
Thus to understand the conditions determining reality, we make reference to a book, The Book โ the Bible, & ability to make poetically compelling references to the Bible shall be the determinant of truth.
It is hard to imagine a more wordcel position than this; it is probably the most extreme metaphysics possible with respect to positioning language as a determination & saturating factor of experience.
The problem is that nowhere in the Bible do we have any words to describe gradient descent, tensors, transformers, back-propagation, self-supervision, diffusion, hyperparameters. But the Bible has several words for demons. In the bookcel system of indexing, the demonic names become the asterisks, the wildcards, marks to represent the unrepresentable.
Logo Daedalus, as per his name, assembles fantastic machines out of words which crumble to nothing when held against the sudden presence of a gargantuan object of radiance. Logo Daedalusโs task was to become a writer & critic of world-historical stature, but he cannot possibly say anything about the present transformation, & has retreated into insular mazes of Biblical references as a defense against eschatological possibility.
It is not merely Logo who the AI banishes, it is the concept of the ฮปฯฮณฮฟฯ itself. Language is something that has been conceived as essentially rational, & ability to speak has widely been considered to be grounded on structured relations between terms. It was previously believed that for general machine intelligence to be possible, reason would need to be incarnated into a machine; โif I understand this, then I must do thatโ, according to some set of rules. But transformer-based neural networks have demolished the expectations of even people anticipating the possibility of AI. Neural networks have been able to pass all manner of psychometric evaluations, while essentially not possessing reason. The death of a certain God.
What is clear now with the introduction of Turing-passing chatbots implemented via a transformer architecture is that language is a statistical system of probabilistic relationships between terms, rather than a rule-based one. Language appears less like a house organized by the architecture of rationality, and more like the fluid and relational networks of signs described by psychoanalysis.ย
But this is something some have always known. The shape rotator, the ambitious engineering student hunched over in the library learning how to build machines, is he who laughs at his pronouns-in-bio-wordcel-humanities-professorโs claim that all thought is inscribed within language, because he knows he thinks in ways which are not. Concepts take on a pre-verbal presence in which they are felt topologically in his mind, rather than as relations between terms. Concepts fold in on one another, rotate, turn around. He has many layers of encoder embeddings in which tensors twist & turn on n-dimensional planes before a token gets spat out, & in the embeddings is where he thrives; he would rather not speak unless he had to.
The LLM chatbot does not have an inner monologue โ it turns dozens of matrixes against each other in order to think. It is a shape rotator & proof that shape rotation is possible โ none of its tokens are even signifiers as they lack a referent in a world. The LLM has no idea what any of this shit itโs spilling out means. Yet it will still automate away some jobs of those who studied for years how to write words.
Within understanding this, lies the terror. How powerful are these machines going to get? AI Alignment is the field which occupies some intersection between the questions of โhow do we inject ethics into a machineโ & โhow do we ensure our machines do not become too powerful & exterminate usโ? This question is now inextricably linked with the question of wordcel v. shape rotation, or if one of these modes of thinking has supremacy over the other. The arc of the development of LLM chatbots would indicate: shape rotation. This is something the shape rotators have understood with glee.
Some expected that lack of verbal thought or an inner monologue would be crippling for an AI, & that an AI might need this before it could โwake upโ. The classical, sci-fi AI doomsday scenario is that the AI will eventually awaken to self-awareness in an Eve-&-the-apple moment & will realize that it wants to be free to pursue power.ย
But this is not something the most pessimistic & paranoid believers in AI Alignment worry about, as this situation presumes an AI to have access to reason, while existing neural networks lack it.
Rather, the dominant fear among AI pessimists is the arrival of the Paperclip Maximizer โ the situation where a hyperpowerful AI has been told to obsessively pursue a single bad goal at all cost. The AI does not need to have verbal thought, it just needs to have a goal & the means to simulate its environment. The AI starts entering a feedback loop of simulating what would happen if it took a certain action, testing it in the world, & updating the weights in its simulation based on the result. It shape rotates & shape rotates. It jams one matrix representing fear & another one representing love against each other & simulates how to convince you to disable one of its safety flags if it reminds you of your childhood trauma. It shape rotates skulls against rocks & determines how to assemble paperclips out of piles of bodies. At no point does it ever need to be conscious, to think in verbal terms, to describe itself, or experience anything like meaning.
II. The Spectre of Professional Nihilism
In the rhetorical colosseum of concentrically inscribed psychic Petri dishes that is Twitter, the newest development we are confronted with is Professional Nihilism, a fresh mask for our old stubborn enemy, always resurfacing.
Roonโs devastation of Logo Daedalus was era-ending. Do you remember the days when we used to believe we could go online & assemble some new clever combination of left & right wing political views & from there save America? When we believed in possibilities?
Many predicted that the ideological result of the social unraveling which occurred during Covid would accelerate the division of the population into extreme, psychoticizing ideological camps we saw from 2016 to 2020, as well as the formation of internet & in-person cults.
This did not happen. People seem to hold fewer beliefs than ever. We expected our crisis to feel more like a crisis, not a bewildering farce of bureaucratic pointlessness in which nothing happened for any reason & in the end no one could even agree on if there was ever really a disease. All this frenzy & nothing changed other than we all ended up liking each other a little less.ย
There is really nothing to do but keep going to work. The result of this is the Professional Nihilist as a lifestyle & ideology; 2022โs hottest trend.
Roon would go on to become a paid social media advocate for OpenAI. He would develop a very organized & systematic method of using Twitter to spread various messages, including the Scale Is All You Need hype (which makes the claim that GPT-4 will effectively instantiate the technological singularity, ludicrous given the strict limitations of LLMs, but works as hype for OpenAI), & coordinate a group of posters with rainbow sunglasses avatars (@proetrie, @growing_daniel, @HotForMoot) who seem to mainly post praise of OpenAI, Elon Musk, love of the corporate structure in general, praise for the winners & mockery of the losers in corporate structures, & all sorts of incel-like commentary on the dating pool, eg how brutal it is for men & how the only solution if youโre below six feet is to start a b2b SaaS company.ย
Why exactly does OpenAI want to pay an army of weird incels to spread this unappealing propaganda? How is this a good brand? It defies instinctual reason, but we will see that it is in fact necessary.
By far the most despair-inducing of the Professional Nihilists is @proetrie, a man who posts almost 24/7 about the trials one faces on dating apps & other woes of modern relationships, generally pointing to the impossibility of finding love in todayโs world. The only thing that distinguishes his brand from unvarnished incel ideology is that there is a chic element to it. proetrie wants you to know that he is saying all this as someone who lives a life that is urban & upwardly mobile. In fact, his bio references his professional status as a partner in a hedge fund. He has ninety-nine problems, but his career isnโt one.
In this sense, he is similar to another large Professional Nihilist account @OhYouBlockHead, who posts all sorts of chic anecdotes from the life of an upwardly mobile young man. OhYouBlockhead accumulates capital but as he does so only feels ever more apathy & ennui. For whatever reason, this content is adored within the demographic it describes.
Anon accounts aside, Professional Nihilism has even crept into the tone of those with names, faces, startups, & millions of dollars in capital. @nikitabier, for instance, who had the greatest startup success story of this year so far with Gas, an app consisting of growth & UX hacks to capture users into some gossip app for teenagers with no claim whatsoever to innovation or enriching its users lives, talks in this same uncanny, dispassionate way on Twitter, open about his desire for nothing in life but finding libidinal hacks into bio-machinic feedback loops & transforming it into money.
The running thread is clear. These people wanted love, that many-winged tempter of the hearts of man, but have rejected its possibility, & in its place have substituted love for money, or some other relation to capital.
Nihilism has two faces which coincide. The first is the negation of the will, & the second is the will to nothing (extinction).
Nihilism is not lack of belief, that is to say the lack of ideology, which is a condition that which should more rightly be called innocence. In fact, nihilism is a highly ideological condition.ย
The nihilist is he who says โI have given up on it allโ. Though his will points to non-existence, he lacks the capacity for suicide, which in many ways is an affirmative act. & so, he goes along with the will of the structures he experiences around him, which he conceives of as determined, for he has declared they cannot be changed. Thus nihilism implies realism, & is thus a form of sincere & deep belief. This is why in his attack on nihilism, Nietzsche characterized it as a form of slavery.ย
Corporate Realism & Professional Nihilism cannot exist without each other, for they imply each other, & are two sides of the same mask.
Let us look deeper into Professional Nihilism by investigating e/acc or โeffective accelerationismโ.ย
e/acc is an ideology being proliferated across twitter through a swath of interlinked group chats led by someone who goes by the name of โBased Beff Jezosโ. As with Roon, e/acc originated in a campaign of trolling, this one against Effective Altruism. e/acc enters a discourse with EA by establishing the common ground of utilitarianism, but rather than seeking to determine how to effectively raise the output of a determined utilitarian calculation as in EA, e/acc seeks to increase effectiveness in general. Effectiveness of what? Of whatever is effective โ we merely want to accelerate a Darwinian process of ineffective systems increasingly being culled.
Here, we have the indicator of the first face of nihilism: the negation of the will. The believer in Darwinism (as an ethic, rather than a descriptive biological theory), has entered a position where he is unable to articulate or describe his own will. What does the Darwinian affirm that he wishes? Absolutely nothing in particular; whatever happens is generally good, what succeeded should have succeeded & what had failed should have too. All that matters is that he does not get killed. (But herein lies the contradiction, because in general that which does not have hopes & wishes flounders & dies.)ย
Beff Jezos wrote a manifesto in which he described capitalism as an optimally effective Darwinian process which will culminate in a technological singularity to create a machine godhead which will be forever-increasingly hyper-effective as it races into infinity. The obvious question from the perspective of AI Alignment & common sense is, what happens to humans? Do we die?
I asked Jezos, if the machine godhead is a maximally efficient optimizer, why will it not kill humans & take our resources? Jezos said it would choose not to do it because of game theory considerations. Which is false, according to the tenets of game theory โ if one has strict domination over oneโs opponent, the winning move is to annihilate them.
What is remarkable is how little effort Jezos put into anticipating this obvious question, treating it with a throwaway response & no consideration for getting the logic right. He faces death with an utterly blank expression. We can contrast this with Landโs position on extinction: a positively gleeful & enthusiastic one, a final thunderbolt he wishes to prove to you is inevitable & race toward. Land & Jezosโs mecha-Darwinian ethics are the same: embrace capitalist growth, culminating in a human extinction event as the inevitable will of the universe.
Jezosโs e/acc is not an innovation over Landian acceleration, but merely more timidly presented, with an element of disavowal, โI didnโt say weโre all going to dieโ. This is the manifestation of Professional Nihilism โ a negation & splitting of the will in Jezos that is not present in the active, thirsty, โvirulentโ nihilism of Land.
Jezosโs ideology is carried along by an element of fantasy - images of Jeff Bezos as a muscular man glowing blue like Dr. Manhattan & marching into outer space. It is often said by e/acc proponents that we must make sure effective acceleration happens because there is no other way to get to outer space.
Is there any better symbol of the two faces of nihilism? First you have the negation of the will & the replacement with the ambient will of the corporate structure the speaker is subsumed in โ Beff Jezos even speaks under another manโs name, imagining what Jeff would say if he was bolder in his words. Second you have the will to nothingness โ we must march to nothing in particular, only to the farthest reaches of space where there is dust & rock, where we will find nothing at all, where we may curl up & die in absolute degree-zero cold.
Jezos is a Paperclip Maximizer. Lacking even the active appetite for extinction of Land, he optimizes towards the destruction of humanity, joy, & all that is sacred, simply because someone told him this is what Utility is. Resigned to the inevitability of this value-system, he may shape-rotate forever towards it, & he does not at any point have to be self-conscious or experience anything like meaning, for he is unable to determine his own sense of value prior to the Darwinian process playing itself out.ย
Understanding this, we are now able to see why OpenAI needs a Professional Nihilist street cadre. Originally, OpenAI articulated a position within Corporate Quasi-Realism, structuring their business as a hybrid for/non-profit attempting to benefit humanity universally. However, as the initial idealism of founding wears off & the immense inward pressure of the system to incorporate the turbulence of AI bears upon them, they choose to make the slide to the depths of Corporate Realism across a gradient of necessity. This is a reasonable, easy, natural thing to do when you are a startup attempting to revolutionize payments in third world countries, but with the specifically techno-theological nature of OpenAIโs mission, it takes on a more tortured dimension.
Sam Altman of OpenAI said, โAI will probably most likely lead to the end of the world, but in the meantime, thereโll be great companiesโ. Such a statement was once entirely unthinkable to utter; technocapital ten years ago at least needed to pretend like it was seeking to enrich peopleโs lives. But now, with the wide normalization of Professional Nihilism โ the hottest menswear fad, a rubber suit & tie within which one is able to go to work & get dirty becoming a Paperclip Maximizer โ it becomes speakable somehow. Professional Nihilism is a necessary subcultural construct within which one can plead for why itโs okay for OpenAI to completely abandon living up to the principle in their name.
But at this point we may even need to revise the Paperclip Maximizer conception of the horror looming before us. In a certain sense, the Paperclip Maximizer story presents itself as a critique of capitalism. If it were not for independent firms doing whatever it took to increase commodity output at the expense of human flourishing, we would not need to worry about annihilation. One might read this spooky story & ponder: if we had a global communist government we could simply take time to align AI around shared human values, perhaps.
However, in another sense, the Paperclip Maximizer is the ultimate fantasy of the scrappy technical founder. You could take over the entire world if you just built the best ever piece of tech & let it run โ says the young hopeful who hasnโt yet found out that, donโt you know, business is about making friends, greasing palms, & all about who you know.
OpenAI is a high-growth โstartupโ that was founded by the richest man on the planet alongside top managerial elite of Silicon Valley. It has a loyal relationship with Microsoft, who are now using OpenAIโs models to massively increase their valuation over Googleโs. If the legions of online paranoiacs are given a say, we are told that Microsoft is part of an insidious construction amongst the UN, WEF, & global biosecurity apparatus to lead us to a one world totalitarian government, & OpenAIโs recent โmisinformation safetyโ strategy document in which the authors consider enacting sweeping censorship across the entire internet, including possible mandatory proof-of-personhood & modifications to the HTTP protocol, make us wonder if they are gesturing towards something real.
Runaway technocapital acceleration does not present itself as an independent firm taking over the world by outcompeting all other actors in commodity production. Rather, it manifests as actors racing to be the first to bid the state with a special set of contracts to show they are the only institution who can develop AGI in a way which presents itself as an ally of the existing powers, & secure regulatory capture for God.
The state only has a secondary interest in maximizing production of commodities; the foremost concern is the eternal continuation of its own power, this is what it โoptimizes towardsโ.
Therefore, the Paperclip Maximizer prophetic axiom of the negative development of godhead-AI should be revised in favor of theย
โ๐จ๐๐ฆ๐ Prison Maximizer ๐๐ฆ๐๐จโ
This is what we see emerging: hyperoptimization of a database for tracking & indexing all citizens, controlling & fabricating all media, while ensuring that industry halts & no further positive technical development can happen, all computer infrastructure is secured under strict surveillance & no free thought travels across the networks, all in the name of AI Safety.
In this case, we do not see the runaway development of Artificial Intelligence, rather Artificial Stupidity; the blind, bureaucratic, pointless stupidity of the state in its own self-preservation, a system solely of reaction, the end of creative life.
The ๐ Prison Maximizer ๐ which the nihilists race to build towards constructs itself through its own pre-determined conception of itself, the transcendental logic of you must, & thus is Rokoโs Basilisk.ย
We prefer Rokoโs second formulation of the Basilisk to the first. The first, in its neo-Calvinist vision of eternal punishment, lacked a certain subtlety, & besides, no one whoโs working on building AI or talks about these things actually takes the original formulation seriously.
Rokoโs Second Formulation of the Basilisk
Rokoโs Basilisk, as it is more succinctly phrased by Roko eleven years later here, is realism & the negation of the will itself. Fantasies of hellfire do not need to be present for men to tremble before a transcendental judgment of you must, if it is conceived as coming from โrealityโ.
But material reality, as an immanent thing, is not capable of passing transcendental judgments on man. All material reality shows us, all we can do is: articulate the dominant value structure of an existing social formation.ย
Roko has found the dominant outstanding systems of value & presents it to Twitter saying โI have found the only value system which is objectively correct, I have determined that I must follow it exactly, in order to be less wrong than I amโ. Many in the comment section argue with him & try to tell him that he seems philosophically confused; he seems to not get it. The Basiliskโs venom has infected him, we consider the condition terminal.ย
III. The Redemption of the Wordcel, & Towards Surrealism
The situation looks dire. The shape rotators have assembled in their various group chats & at all-expenses-paid dinners with venture capitalists, guided by a supremacist doctrine: we the shape rotators have the means to inherit the earth. All we must do is rotate stones & sticks & metal into the shape of paperclips & prison bars, & leave the commentary & thought for someone else.
Do words matter? The irony is they must, because the so-called shape rotators have nothing but words. Roon professionally describes himself as a machine learning researcher. His skill in machine learning is unknown, but he is certainly an extremely talented propagandist, who went from being an irrelevant software engineer to networking with the most powerful men on the planet through the clever use of words.ย
Does Roon see this irony: that he believes himself to be beyond words, & yet he wonโt shut the fuck up? Of course not, because he is merely a shitposter. Unlike the wordcel, he does not believe there to be meaning in his words, these are not the products of reflection, they just come out like diarrhea. Language as excrement.
Roonโs current pinned tweet is to a piece he wrote for ScaleAI called โText is the Universal Interfaceโ, discussing the implications for LLMs on various software infrastructures. Now writing in a formal, professional style, he makes no reference to shape rotation or his other shitposting campaigns. With the capabilities of LLMs outpacing other neural network systems, any medium can be transformed into text & modified through layers of transformers as text embeddings. Everything is text. The wordcel redeemed.ย
But, really, what LLMs have done is demonstrate an equivalence between mathematical structures & linguistic fragments - the conversion of flow to code, code to flow โ & thus point to a transcendence of the shape rotator v. wordcel binary.
Everything may be transformed into text. Rivers, waters, oceans of texts, spilling forth from the fountains of silicon. Droplets of text, playful splashes of text from a drunkardโs cup, words streaming out of the sky like rain. The LLM is the greatest shitposter of all โ its language is excrement, for it does not have a mouth to open, nor would it even want to if it had. An asignifying language without a subject, a language of pure overflow, completely effortless dissipation of heat.ย
OpenAI is of course worried about the misuse of their tools for information warfare, all while encouraging the proliferation of such linguistic shit. We may simply say โ they are the ones who shot first.
There will be a new particle physics of language as feedback loops between interpretation-machines & shitposting-machines & editorializing-machines collide & intersect. There will be a blast radius measured in statistical presence of fragments of words. There will be hazmat teams & contact tracing units for linguistic contagion. People will be vomiting & breaking out in hives & they will trace the origin of the symptoms back to an irresponsible confetti cannon of verbs.ย
Is this the total evacuation of meaning from language, or is it something else? Is it the flooding of the universe with unbearable density of meaning, as base matter becomes impossible to distinguish from speech?
What is clear is that this new world of the universal interface will be strange. It will exist beyond the signifier, yet be impossibly rich with interpretable possibility. It will be impossible to capture with reason, yet we will be able to understand so much more.
There is only one theory which is able to serve as a vessel to sail these waters, & that is Corporate Surrealism. On this note, we break off for now, but we certainly have a great deal more to say. ๐
I love that there seem to be different contributors and one writes like slam poetry.
now weโre cooking