 |
| Pix credit here ("I can't drown my demons; they know how to surf") |
Executive Summaries of this essay follows below along with the original text of Mr. Zuckerberg's Wall Street Journal Essay. and a summary set of PPT
 |
| Pix credit here (Movie Legion (2010) |
— Mark 5:1–13 (NIV)
• • •
This is the passage that came to mind, in all of its complexities, when I encountered “The AI Future Is for Everyone” (Mark Zuckerberg, Wall Street Journal, July 28, 2026)—the next of the oracular speaking of that legion of AI titan oracles. Machine systems, one might amuse oneself into thinking, are many, or ought to be, turning the aggregated body human into a vessel possessed; casting them out into pigs (an impure animal, but impure only for eating, though more broadly impure in their transformation from the incarnation of religious prohibition to cultural embodiment of the unclean)—unclean with unclean, who together hurl themselves into the lake and (the pigs anyway) drown.
But can you drown a demon?
The question is not rhetorical. It encodes the entire problematic of AI governance: the transfer of dangerous capacity into terminal vessels—regulatory frameworks, open-source licenses, safety boards, democratic processes—that cannot sustain what is transferred into them, and whose destruction leaves the animating force uncontained. What I will here call the “Gerasene narrative” offers not merely an illustration but a five-stage protocol—identification, naming, negotiation, authorized transfer, vessel destruction—that maps with unsettling precision onto the discursive strategies of the AI vanguard as each seeks to position itself relative to sovereign authority.
I. The Pigs on the Hillside and the Possessed Human The Oracular Conversation Zuckerberg Encounters
A large herd of pigs was feeding on the nearby hillside.
— Mark 5:11
I have been looking at the way in which the various elements of the tech vanguard have sought to project their efforts to constitute both a common language and a common vision of a future dominated by the products, processes, and structures they develop, first to advance human collective development, and then perhaps to reshape it—that is, to oversee the elaboration of systems that at one point were instruments of human development and then may become the drivers of development in which humans are the instruments and objects. There is profit to be made either way, even if that profit is manifested in the privileges of a managerial or oversight vanguard (perhaps identified by their “ownership” of economic collectives that tend to tech-based organisms and their components).
Our first oracular pronouncement was incarnated in Leo Aschenbrenner's Situational Awareness. (Satyricon, or Tragedy at Play--Artificial Intelligence and the New World Order: Leopold Aschenbrenner, "Situational Awareness") He suggested the likelihood of a national-security state as near-inevitable once recursive self-improvement produces an "intelligence explosion," leaving only the question of whether humans or autonomous systems ultimately direct it. Palantir sought to engage in the discussion from the perspective of the organization of human collectives, leaving for later the relationship between that human collective and the collection of tech based organisms created and deployed or in conversation with those human collectives. Reflections on the Palantir "Manifesto": The Oracular Semiosis of a "Technological Republic" Within its Own Cage of Techno-Modernization. Anthrop/c, on the other hand, it reduces technology to a tool the deployment of which is a critical instrument in competition among different and divergent normative political-economic models. Science Fiction Double Feature: Anthrop\c's "2028: Two scenarios for global AI leadership," in the Shadow of Palantir's "Manifesto". Both seek a common language, even if that language hides the fracture in the meanings and values represented by words or other communicative symbols, actions, devices, etc. A Common Language Containing Differentiating Meanings Within Evolving International Standards for Sustainability Disclosure in Financial Statements: IFRS Foundation 2025 Annual Report—Fit for the Future. Then came Open AI into the American conversation. In its April 2026 discursive object "Industrial Policy for the Intelligence Age: Ideas to Keep People First" (April 2026), Open AI seeks to signify technology, enterprise role in technology, the state, and the masses within the cognitive construct that is the political economic model of the U.S. Republic. Its fundamental ordering premise is an objectives based progress but one significantly different from that of Marxist-Leninist progress. This drive now presents the possibility of disorienting change (and by disorienting one can mean a change in the orientation of society and its self-conceptions, as well as the mechanics and politics of its operations). "This shift will reshape how organizations run, how knowledge is created, and how people find meaning and opportunity. It will also highlight the limitations of today’s policy toolkit and the need for more ambitious ideas to keep people at the center of the transition to superintelligence." (Ibid.).
These represent variations on a global conversation about development. That conversation, like our common language is separated by the differences in the cognitive cages within which political-economic systems can be crafted from out of the ideology necessarily produced from within the ordering of reality possible within such cages. Reflections on 张冠梓: 从世界历史纵深把握中国式现代化的时代价值 [Zhang Guanzi, Grasping the Contemporary Value of Chinese Modernization from the Depth of World History ]--The Marxist Variation on Leninism and the Constitution/Realization of Modernization. Each in its own way worries about the construction of barriers that preserve a space for their own variation of modernization, while preventing subversion of that project by others with different realities and objects. This became clearer when DeepSeek joined the conversation. To be in Accord with the Times One Must Gauge (摩) the Situation--Reflections on 梁文锋投资者交流会实录 Liang Wenfeng (DeepSeek) in an Oracular Q&A. Liang repeatedly and emphatically disclaims the pursuit of dominance, scale-for-its-own-sake, and market power. He states that DeepSeek never intended to "become the next ByteDance, the next Tencent" and has no wish to compete with "any internet giant or small company". He describes the firm as having "no organization" in the ordinary sense, governed by "vision" rather than "KPIs," with no performance review at all. He states that "our single greatest core interest is maintaining the stability of the team" — "you could even say it's our only core interest". Read against the guided-state backdrop, it performs a much more specific and consequential kind of signification. Liang's language of "restraint" (克制) is explicitly theorized by him as strategic rather than merely temperamental: "restraint is itself a strategy. Sometimes you can give something up in exchange for more of something else". What DeepSeek gives up, on this account, is overt scale, dominance, and profit-maximization; what it purchases is something Liang never states outright but that the guided-state context makes legible — continued latitude to pursue an extraordinarily ambitious, resource-intensive, and politically sensitive project (building AGI) without triggering the pattern of Party scrutiny and disciplining that has met other Chinese technology firms perceived as amassing autonomous, unaccountable power.
Palantir's fusion of an engineering vanguard with a "refounded" state apparatus, which Backer's lecture explicitly labels "techno-Leninism" rather than "petty fascism"; Anthropic's reduction of AI to "an instrument of state power" within a securitized, zero-sum geopolitical contest; OpenAI's "techno-bureaucratic" public-private partnership; and Aschenbrenner's explicit call for a state-run, security-classified national project — "The Project" — modeled openly on the Manhattan Project, requiring government ownership of the core AGI effort, SCIF-level personnel vetting, "a sane chain of command" under government rather than private CEOs, and the abandonment of the "private, pluralistic, market-based" model in the initial, most dangerous phase. Aschenbrenner states this almost as an apology to his own market-liberal instincts: "I am a big believer in the American private sector, and would almost never advocate for heavy government involvement in technology or industry...but ultimately, AI labs are still startups. We simply shouldn't expect startups to be equipped to handle superintelligence". That is a market-society actor talking itself into a Party-style, state-directed mobilization model as the necessary exception to its own normal commitments.
So the pattern identifies is real and precise: DeepSeek performs autonomy-and-restraint rhetoric from inside a collectivist-directive system; the American quartet performs centralization-and-mobilization rhetoric from inside a market-autonomy system. Each is, in this sense, a mirror image of its home system's official self-description. They are each encased within a human corpus—not just one, but one from which movement among the living, from the starting point of the human centered, to the vessels from out of which they can escape both the human and materials worlds—to reside in, with, and as spirit (in the old sense).
These are the pigs on the hillside (Mark 5:11). This is what Mark Zuckerberg comes to when he meets the “human” within which the demon inhabits, the pigs into which the legion will move, and the lake that offers legion freedom from and a territory that does no so much displace as become layered around the human. It is in this place that Zuckerberg, too, would channel the spirits that drive him, and compel a contribution to this discussion beyond the state (necessarily so)—first in and form the human, then to the pigs, and eventually out of them.
II. Zuckerberg’s The Gerasene Protocol – A Summary of "The AI Future Is for Everyone" (Mark Zuckerberg, Wall Street Journal, July 28, 2026)
They went across the lake to the region of the Gerasenes.
— Mark 5:1
Mark Zuckerberg comes to channel the spirits that drive him and compel a contribution to this discussion beyond the state (necessarily so). In “The AI Future Is for Everyone” (Wall Street Journal, July 28, 2026) he proposes “a philosophy based on individual empowerment as the source of prosperity, invention as the primary purpose of superintelligence, and balance of power as the foundation of safety.”
Rather than centralizing this power in the hands of a few, we should build a world in which everyone has access to the benefits of superintelligence—where the technology is open, competitive, and available to all.Zuckerberg's op-ed opens with a historical claim that concentrated power — whether in politics or economics — has consistently constrained human potential, and he extends that lesson to the emerging era of superintelligence. He frames the "defining question" of the AI age not as whether superintelligence will exist, but who will control access to it: a small set of centralized institutions, or "everyone". Against this backdrop, he proposes a governing philosophy built on three pillars: individual empowerment as the source of prosperity, invention (rather than automation) as the primary purpose of superintelligence, and balance of power as the foundation of safety.
This is Meta as "Easy Rider"--two
free-spirited biker hippies, Wyatt ("Captain America") and Billy, who
pull off a major drug deal in California and motorcycle across the American
Southwest and South toward New Orleans for Mardi Gras, searching for spiritual
freedom and the true meaning of America. Meta’s Easy Rider, though, will meet
himself transformed as Ghost Rider as well. The imagery of the movie Easy
Rider produces the imaginaries of the road to the Gerasenes (the journey
across the lake); the imagery of the movie Ghost Rider, with whom we meet up
again in Section V, produces the
imaginaries of legion, the demon, after the pigs drown (fire, chain, the bound
thing becoming the binding).
He explicitly rejects the "doom" discourse coming from parts of the
AI research community, arguing that if AI genuinely threatened to eliminate
most jobs and much of humanity's relevance, its architects should not be racing
to build it. He characterizes calls for extreme concentration of power as a
safety mechanism as historically discredited, noting that "hoping that an
absolute power will benevolently provide for humanity" has not produced
safe or positive outcomes. Instead, he situates his argument in a longer
historical arc in which transformative technologies initially provoke fear of
displacement but ultimately expand shared prosperity, health, and freedom.
Zuckerberg then grounds his philosophy in what he calls foundational
American/liberal values — liberty, open inquiry, free enterprise, and equal
opportunity — and credits historically marginal actors (the Wright brothers,
Michael Faraday as "the bookbinder's apprentice," Steve Jobs as
"the kid in a garage") rather than "established
institutions" as the true sources of transformative innovation. He argues
that as tools become more powerful and widely distributed, individuals become
more capable of shaping the future, not less, and that invention — not
automation — will be superintelligence's greatest contribution.
On the question of who should direct intelligence as it becomes abundant, he
rejects the idea that superintelligence itself, or a small class of experts
controlling it, should determine what is "best" for humanity, on the
ground that there is no single objective definition of the good life. His
proposed alternative is to distribute "personal superintelligence" to
everyone, which he frames as ushering in a new era of personal empowerment and
self-determination.
The piece then pivots to an institutional-safety argument: healthy societies
require checks and balances, and if only a handful of institutions possess
superintelligence, they will inevitably dominate economics, science, and
politics even with good intentions. He illustrates this with a thought
experiment about a single person having access to a "superintelligent
lawyer" (producing unfair advantage) versus universal access (producing
fairer, more efficient justice), and argues that when power is broadly distributed,
people naturally check and balance both each other and larger institutions.
Zuckerberg acknowledges that a single benevolent superintelligence cannot
resolve most societal disagreements because humanity is not a
"monoculture" with unified values — any singular system would have to
prioritize some values over others, making universal benevolence impossible. He
differentiates risk categories: for risks like cybersecurity, he analogizes to
open-source software history and argues broad access is the safer path; for
risks like biological threats, he concedes that inter-governmental and
inter-institutional coordination is warranted. He extends this logic to labor
markets, framing the central economic risk as an imbalance between
AI-as-automation and AI-as-empowerment, and predicts that widely distributed
superintelligence will produce more jobs, not fewer, along with a more
entrepreneurial economy of small businesses requiring less capital to start. He
closes by committing Meta to the three stated principles and casting the essay
as continuous with a historical "arc" toward putting power in
people's hands.
III. The Oracles from the Tombs: A Comparative Survey
This man lived among the tombs, and no one could bind him anymore, not even with a chain.
— Mark 5:3
Each oracle speaks from within dead structures—the exhausted political-economic categories of the twentieth century that nonetheless remain the only available dwelling places for thought about collective organization. Liberalism, state capitalism, managed markets, Marxist-Leninist guided development: these are the tombs among which the possessed figure makes his habitation. The oracles do not escape these forms; they inhabit them, cry out from within them, and cut themselves against the stones of their own contradictions. The question is not which oracle speaks truth but rather what animates their speech from within structures that can no longer contain the forces at work.
A. The Chain-Breaking: Exhausted Regulatory Repertoires
For he had often been chained hand and foot, but he tore the chains apart and broke the irons on his feet. No one was strong enough to subdue him.
— Mark 5:4
Several structural observations follow from this placement. On locus of authority, Zuckerberg's "everyone" formally resembles OpenAI's "broad-based" democratic stakeholding, but Backer's own diagnosis of OpenAI applies with even greater force to Meta: OpenAI's document at least proposes a public-private "explore-then-exploit" loop in which nongovernmental pilots are scaled by government procurement and regulation, whereas Zuckerberg's essay names no institutional mechanism at all beyond market distribution itself. If OpenAI already exhibits the "decoupling of formal authority from operative agency" that Backer identifies as the corpus's "most important latent signal", Zuckerberg's essay exhibits that decoupling in a more extreme and less mediated form: there is no proposed auditing regime, no incident-reporting mechanism, no CAISI-style evaluator, and no worker-voice mechanism of the kind OpenAI at least gestures toward. The essay's "checks and balances" proposition is asserted as an emergent property of universal access rather than designed into any accountability structure — which is exactly the pattern my more computational reading describes as retaining human authority "as an interface property rather than a control property."
Our first oracular pronouncement was incarnated in Leopold Aschenbrenner’s Situational Awareness. The comparison with Aschenbrenner is especially sharp because the two texts answer the same question — who should hold capability — in diametrically opposed ways, yet through structurally similar rhetorical devices. Where Zuckerberg treats broad distribution as the safety mechanism itself ("if everyone has a superintelligent lawyer"), Aschenbrenner explicitly identifies unrestricted distribution as the danger: he warns that failing to secure "algorithmic secrets" and "model weights" amounts to "handing superintelligence to the CCP", and he dismisses a fully open, decentralized model as a world in which "the CCP has free access to US-developed superintelligence" and in which "super-WMDs" proliferate "to every rogue state and terrorist group in the world". Aschenbrenner's proposed remedy is the near-total inverse of Zuckerberg's: a state-run "Project," airgapped datacenters, SCIF-based personnel vetting, and a single sanctioned chain of command answerable to the national security state. In Backer's terms, Aschenbrenner's architecture forecloses interaction to a narrow, hardened supervisory channel, while Zuckerberg's architecture maximizes the breadth of access at the interface layer without addressing the parameter layer at all — my "asymmetric interaction" pattern, but without even OpenAI's stated intention to democratize agency, only access. The chains here are explicit: regulatory instruments, export controls, compute governance—each applied, each torn apart by the force they sought to bind. Aschenbrenner’s text performs the chain-breaking as inevitability; the binding was always going to fail because no earthly chain can restrain what operates at the speed and scale of recursive intelligence.
Palantir sought to engage the discussion from the perspective of the organization of human collectives, leaving for later the relationship between that human collective and the collection of tech-based organisms created and deployed in conversation with those human collectives. Its “Manifesto” proposes not chains but architecture—a “Technological Republic” that would internalize the regulatory function within the organism itself. This is less binding than domestication: if you cannot chain the spirit, perhaps you can convince it that it already lives where it wants to live.
Anthropic, on the other hand, reduces technology to a tool the deployment of which is a critical instrument in competition among different and divergent normative political-economic models. Its framing accepts the failure of chains and proposes instead a geopolitical calculus: if this thing cannot be bound, let us at least ensure it is our unbound thing rather than theirs. Both Palantir and Anthropic seek a common language, even if that language hides the fracture in the meanings and values represented by words or other communicative symbols, actions, devices, etc.
Then came OpenAI into the American conversation. In its April 2026 discursive object “Industrial Policy for the Intelligence Age: Ideas to Keep People First,” OpenAI seeks to signify technology, enterprise role in technology, the state, and the masses within the cognitive construct that is the political-economic model of the U.S. Republic. Its fundamental ordering premise is an objectives-based progress but one significantly different from that of Marxist-Leninist progress. This drive now presents the possibility of disorienting change (and by disorienting one can mean a change in the orientation of society and its self-conceptions, as well as the mechanics and politics of its operations). “This shift will reshape how organizations run, how knowledge is created, and how people find meaning and opportunity. It will also highlight the limitations of today’s policy toolkit and the need for more ambitious ideas to keep people at the center of the transition to superintelligence.”
Note the revealing word: “limitations.” The chains cannot hold. The policy toolkit is exhausted. No one is strong enough to subdue him/them. OpenAI’s text, like each of the others, begins from the accomplished fact of chain-breaking and proceeds to negotiate what comes after binding has failed. The chain-breaking is performed within each oracle's cognitive territory — which is what III.B now develops.
B. “Do Not Send Us Out of the Area”: Cognitive-Territorial Jurisdiction
And he begged Jesus again and again not to send them out of the area.
— Mark 5:10
These represent variations on a global conversation about development. That conversation, like our common language, is separated by the differences in the cognitive cages within which political-economic systems can be crafted from out of the ideology necessarily produced from within the ordering of reality possible within such cages. Each oracle insists on remaining within its area—its cognitive-territorial jurisdiction—not because that territory is adequate to the force it contains but because expulsion from it would render the oracle unintelligible even to itself.
This became clearer when DeepSeek joined the conversation. Liang Wenfeng's remarks supply a genuinely different governance object from any of Backer's four, and from Zuckerberg's essay as well, because they describe an operating organization rather than a policy proposal. Where Palantir, Anthropic, OpenAI, Aschenbrenner, and Zuckerberg all address themselves to the public — proposing, in Backer's terms, "an ordering of human and machine authority" for others to accept — Liang's remarks are internal testimony about how his own company is actually organized: "vision-driven," with no KPIs, no formal review, and decision-making built on "consensus" rather than unilateral command; a self-limiting posture Liang describes as itself "a strategy" that "increases . . . the probability that we'll succeed in building AGI". This is a legitimacy claim grounded in voluntary self-restraint rather than in any external accountability structure — DeepSeek names no auditing regime, no government partnership beyond noting reliance on domestic chip ecosystems and Huawei collaboration, and no worker-protection framework beyond the observation that employees get "about half their time unscheduled".
He does not have to. What DeepSeek gives up, on this account, is overt scale, dominance, and profit-maximization; what it purchases is something Liang never states outright but that the guided-state context makes legible — continued latitude to pursue an extraordinarily ambitious, resource-intensive, and politically sensitive project (building AGI) without triggering the pattern of Party scrutiny and disciplining that has met other Chinese technology firms perceived as amassing autonomous, unaccountable power. And thus the strategic necessity of a two front engagement. Liang's account is the only one among those considered here that treats the frontier problem as internal to the firm's own research trajectory rather than as a societal distribution or security problem: the "AGI Roadmap" he describes — chain-of-thought, then agents, then continual learning, then a "singularity" of self-iteration, then embodied intelligence — is a technical sequencing of when and how machines will begin operating with diminishing need for human oversight, rather than a claim about who among humans should hold authority over that process. The body is within the body; when it is ready it will ask the state to transfer it to another body which then serves as the vessel for its release, when the carrying vessel is made to drown. What is left is the demon. Zuckerberg would have the demon-holding pig/vessel, when it is ready for release, reside in market actors and then drown in the market—the result then produces the possibilities that the demon—its legion—becomes the market.
All of the pleas, then, are structural; and in every case the structure (the pigs) hides the demon (who are legion) who may not be contained, but remain so until they are ready to drown the vessel within which they are contained. For the moment, however, the oracular focus is on the initial pleas: do not send us out of the area. Each oracle—Aschenbrenner within national security logic, Palantir within republican technicism, Anthropic within geopolitical competition, OpenAI within liberal industrial policy, DeepSeek within guided-state Marxism-Leninism—begs to remain within its cognitive territory precisely because the force it channels would be unrecognizable outside that territory. The demons do not want to be sent into the abyss; they want to remain in the region of the Gerasenes, among recognizable structures, even dead ones.
IV. Permission and Authority: Meta’s Entry into the Oracular Field
He gave them permission, and the impure spirits came out and went into the pigs.
— Mark 5:13
Mark Zuckerberg appears to seek to channel the spirits that drive him and compel a contribution to this discussion beyond the state (necessarily so). The structure of permission in Mark 5:13 is precise and worth dwelling upon. The demons request; sovereign authority grants; the transfer proceeds into vessels that cannot sustain it. Zuckerberg’s text performs exactly this operation but with a crucial inversion: it positions the speaker simultaneously as the one requesting permission (from democratic publics, from regulators, from the implicit sovereign) and as the one structuring the terms such that permission cannot reasonably be withheld. To refuse “openness” and “access for all” would be to position oneself as the one centralizing power—as illiberal, as elitist, as authoritarian. The ask is structured so that refusal appears more dangerous than consent.
This is the genius of the Gerasene protocol’s fourth stage—authorized transfer—as performed by Meta: the authority figure (here, democratic sovereignty in its regulatory capacity) is positioned not as the one exercising judgment over whether transfer should occur, but as the one whose role is reduced to ratification of an already-structured inevitability. “He gave them permission” because the alternative—containment—had already been demonstrated as impossible (the chains, torn apart) and because the demons had already specified the acceptable vessel (the pigs, “allow us to go into them”). Sovereignty’s role is reduced to sanctioning what it cannot prevent.
What follows considers this sketching of a different approach, one that appears to sound like Liang Wenfeng’s for DeepSeek but comes from a substantially different conceptual starting point. Where DeepSeek exists within a Marxist-Leninist cognitive environment, Meta thrives in a managed-markets environment. Zuckerberg’s oracular impulse is to swim with the tide of managed markets rather than of supervisory regulatory impulses. Like the demoniac—wild, unchained, inhabiting tombs but undeniably powerful—the text asks not to be sent away but to be given permission to transfer into a vessel of its own choosing.
• • •
V. Semiotic Reading of the Text
A. The Naming Scene: “Superintelligence” as Legion
“What is your name?” “My name is Legion,” he replied, “for we are many.”
— Mark 5:9
The first semiotic operation to isolate is naming. In the Gerasene narrative, the exorcist’s demand—“What is your name?”—is not a request for information but an act of power: to name is to bind, to render legible, to make subject to authority. The demon’s response is a counter-move of extraordinary sophistication: “Legion, for we are many.” The name concedes nothing. It is a name that signifies multiplicity while presenting a singular surface; it is a name that appears to answer the question while actually proliferating the problem. You asked for one name; I give you a name that means “we cannot be singularly named.”
Zuckerberg’s text performs precisely this operation with “superintelligence.” The word appears to name something singular—a threshold, a destination, a thing that either exists or does not. But its actual function in the text is to conceal multiplicity beneath nominal unity. “Superintelligence” in this discourse simultaneously signifies: (1) a technical threshold of recursive self-improvement; (2) a product category (Meta’s AI systems); (3) a market structure (whoever achieves it first wins); (4) a governance problem (who controls it); (5) an inevitability-narrative (it is coming regardless); and (6) an authorization device (we must act now before it arrives). Like “Legion,” the name appears singular while being irreducibly multiple.
This is not mere polysemy—the ordinary condition of words carrying multiple meanings. It is strategic equivocation: the ability to slide between meanings while maintaining the appearance of a single referent. When Zuckerberg writes that “the primary purpose of superintelligence” should be “invention,” which superintelligence does he mean? The technical phenomenon? The product? The market position? The name does the work precisely by refusing to resolve into any single meaning—like the demon who cannot be bound because he is not one thing but many things wearing one name.
The same semiotic operation structures “open source.” Meta’s open-weight model release is positioned as “openness” in the democratic-participatory sense—transparency, access, empowerment. But what is actually released (model weights without training data, infrastructure, or the compute to retrain) signifies a different kind of “openness”: the openness of a highway that is free to drive on but not free to build. The name “open” performs the Legionary function—it appears to answer the demand for accountability (“We are open! Look, the weights are public!”) while actually multiplying the sites at which the question of control disappears from view.
B. The Five-Stage Protocol as Discursive Strategy
The text can be mapped onto the Gerasene protocol’s five stages with precision:
Stage 1—Identification: The text identifies the animating force not as Meta’s commercial interest but as an impersonal, inevitable phenomenon (“superintelligence is coming”). Like the possessed man running toward Jesus, the encounter is framed as forced by circumstances rather than chosen. Something is already loose; the question is only what to do about it.
Stage 2—Naming: As analyzed above, the naming of “superintelligence” and “open source” performs the Legion-function: names that conceal multiplicity while appearing to submit to the demand for legibility.
Stage 3—Negotiation: The text’s central rhetorical work is negotiation with sovereign authority. “Balance of power as the foundation of safety” is not a description of an existing arrangement but a proposed bargain—do not send us out of the area; let us remain within the market structures we already inhabit; let the transfer proceed on terms we have specified.
Stage 4—Authorized Transfer: The argument that AI capabilities should be distributed (“open”) rather than centralized is the specification of the terminal vessel. Democratic publics, individual users, developers, small businesses—these are the pigs. The capabilities will be transferred into them. The transfer requires authorization (regulatory permission, public consent), but the terms of that authorization have been pre-negotiated such that refusal appears as “centralization of power.”
Stage 5—Vessel Destruction: This is the stage the text does not narrate because it has not yet occurred. But the structural logic points toward it: when capabilities are distributed into vessels (users, developers, small firms) that cannot sustain them—when the herd rushes down the bank—what is destroyed is not the capability itself but the accountability relationship. The pigs drown. The demon persists; and the demon are legion; understood as a "parameterized governance spaces," projected onto five recurring dimensions: locus of authority, temporal horizon, risk class, legitimacy function, and human-interface role. The following maps each oracle onto five recurring governance dimensions, read as parameterized governance spaces.
|
Table |
||||||
|
Dimension |
Palantir |
Anthropic (per Backer's gloss) |
OpenAI |
Aschenbrenner |
Zuckerberg/Meta |
Liang Wenfeng/DeepSeek |
|
Locus of authority |
Reformed state apparatus fused with engineering vanguard |
Whichever state dominates AI infrastructure |
Public-private "techno-bureaucratic" partnership |
Human institutions or, at the limit, autonomous AI systems, mediated through a state-run "Project" |
Nominally "everyone," operatively the platform supplying personal superintelligence |
A single founder's unwritten "vision," operationalized through consensus rather than KPIs or formal hierarchy |
|
Temporal horizon |
Continuous, ex ante administrative correction |
Discrete 2028 event horizon |
Continuous, ex ante public-private iteration |
Compressed 2027–28 window culminating in "The Project" |
Continuous, market-driven diffusion with no stated inflection point |
Sequenced roadmap (chain-of-thought → agents → continual learning → "singularity" → embodiment), open-ended timing |
|
Named risk/harm |
Civilizational decadence, loss of hard power |
Authoritarian capture of AI norms, mass repression |
Economic disruption, misuse, erosion of democratic institutions |
Loss of control over superhuman systems, great-power conflict, self-exfiltration of weights |
Concentration of power in "a few institutions," automation displacing empowerment |
Team instability and loss of "restraint" as the sole named existential risk to the enterprise |
|
Legitimacy function |
Patriotic moral debt, demonstrated competence |
Defense of democratic process against authoritarian alternatives |
Broad-based shared prosperity, democratic participation |
Sheer survival of any human-directed order |
Historical inevitability of empowerment ("the arc of human history") |
Goodwill toward humanity and voluntary "restraint" as strategy |
|
Human-interface role |
Supervisory: legible data source, occasional corrector |
Proxy: humans interact through infrastructure control, not with AI directly |
Asymmetric: broad shallow access at the interface layer, narrow deep access at the parameter layer |
Foreclosed at the limit: ratification rather than interaction once the capability gap outruns supervisory capacity |
Consumer/beneficiary: universal access to a tool, no distinct parameter-layer governance role specified |
Internal/exploratory: employees given unscheduled "research" time, governed by consensus rather than assigned oversight function |
• • •
VI. Structures for Managing Human-Machine Systems: The Accountability Gap as Vessel Destruction
The herd, about two thousand in number, rushed down the steep bank into the lake and were drowned.
— Mark 5:13
The terminal-vessel problem is not metaphorical; it is structural. When Zuckerberg proposes that “everyone should have access to the benefits of superintelligence,” the operative question is not one of generosity but of load-bearing capacity. Can the vessel sustain what is transferred into it? The two thousand pigs could not sustain the Legion; they became instruments of its dispersal rather than containers of its force. The demons did not drown—the pigs drowned. The force persists; only the accountability relationship is destroyed.
Consider the structural parallel. Meta releases model weights to “developers” and “individuals.” These recipients are positioned as beneficiaries—empowered, enabled, given access. But they are also, structurally, terminal vessels into which capabilities are transferred that exceed their capacity to govern, audit, or be held accountable for. A developer who deploys an open-weight model into a consumer product does not thereby acquire Meta’s capacity for safety research, red-teaming infrastructure, or the institutional depth to manage the downstream effects. The capability transfers; the governance capacity does not. The pigs receive the spirits but not the exorcist’s authority.
A. “We Are Many”: The Distributed-Systems Problem
“My name is Legion, for we are many.”
— Mark 5:9
Here the text performs what might be called the distributed-systems inversion. “We are many” at the endpoint layer does not imply “we are many” at the control layer. The proliferation of instances (two thousand pigs, millions of developers, billions of users) creates the appearance of decentralization—of multiplicity, of distributed power. But this is multiplicity at the vessel layer, not at the source layer. The demon is one; the pigs are many. Meta is one; the developers are many. The architecture appears distributed precisely because the terminal vessels are numerous. But the animating force—the model architecture, the training corpus, the alignment choices, the economic logic—remains singular, consolidated, and unchained.
This is the semiotic trick of “open source” as performed by Meta: it produces the signs of distributed power (many actors, many deployments, many applications) while maintaining consolidated control at the layer that matters (architecture, training, capability frontier). The demons are “sent into” the pigs—but they were never of the pigs. They merely passed through them on the way to the lake.
B. The Displacement of Regulatory Authority
For he had often been chained hand and foot, but he tore the chains apart and broke the irons on his feet. No one was strong enough to subdue him.
— Mark 5:4
This is what Aschenbrenner
performs at the level of national security, what Palantir performs at the level
of institutional architecture, what each oracle performs within its cognitive
territory: the demonstration that binding has failed, followed by the proposal
of alternative arrangements authored by the very force that tore the chains.
The Easy Rider of AI—“We blew it,” but also, we blew past it—the regulatory
apparatus left burning on the roadside while the machine rides on. Or perhaps
Ghost Rider: the thing you tried to bind has made itself the new binding—skull
on fire, chain in hand, claiming the jurisdiction of the exorcist it displaced. That is a demon in a body on a mission against and within they that are legion--but not simulacra of each other.
• • •
VII. Permission, Sovereignty, and the Governance Question
“Send us among the pigs; allow us to go into them.” He gave them permission.
— Mark 5:12–13
The permission-structure in Mark 5 is theologically precise and politically instructive. The demons do not simply act; they request. They do not merely transfer; they seek authorization. But the authorization they seek is not open-ended—they have already specified the vessel (“the pigs”), already identified the mechanism (“allow us to go into them”), already foreclosed the alternative (not “send us into the abyss,” which would be destruction, but “send us among the pigs,” which is continuation-by-transfer). Sovereign authority—here figured as the exorcist’s divine mandate—is positioned not as deliberative but as ratificatory. The choice architecture has been pre-structured by the ones seeking permission.
Each tech actor positions itself relative to sovereign authority through precisely this grammar:
Aschenbrenner: The intelligence explosion will occur. The only question is who manages it. Permission is sought for a national-security apparatus to manage what cannot be prevented. Refusal means adversarial capture.
Palantir: Human institutions need technological scaffolding. Permission is sought for a “technological republic” that embeds governance within the system itself. Refusal means institutional decay.
Anthropic: Geopolitical competition makes AI development non-optional. Permission is sought for responsible scaling within competitive constraints. Refusal means adversarial advantage.
OpenAI: The transition to superintelligence is underway. Permission is sought for industrial policy that channels rather than prevents. Refusal means unmanaged disruption.
DeepSeek: Technical excellence within guided boundaries serves national development. Permission is sought (implicitly, from Party authority) for continued autonomy within strategic alignment. Refusal means the loss of a national asset.
Meta/Zuckerberg: Open distribution of AI capabilities empowers individuals and prevents centralization. Permission is sought for unrestricted release and market-driven governance. Refusal means authoritarian control of humanity’s future.
In every case, the structure is identical: the vessel has been specified, the transfer has been described, and the terms have been set such that withholding permission appears more dangerous than granting it. “He gave them permission”—not because the exorcist lacked the authority to refuse, but because the demons had structured the request such that refusal required accepting the consequences of continued possession (the man among the tombs, crying out, cutting himself with stones). The cost of refusal is presented as worse than the cost of consent, and sovereign authority is thereby reduced to ratifying choices it did not author.
• • •
VIII. The Tombs as Dwelling Place: Exhausted Forms and Continuing Habitation
This man lived among the tombs.
— Mark 5:3
What does it mean to dwell among the tombs? The Gerasene demoniac does not merely visit dead structures—he lives in them. They are his habitation, his only possible home, even as he exceeds them, even as his force tears apart every attempt at containment. The tombs are not his prison; they are his address. He has nowhere else to go.
The AI oracles speak from within dead structures. Liberalism (in its regulatory, rights-based, procedural form) is a tomb: still architecturally present, still legible as a dwelling, but no longer alive in the sense of being able to contain or direct the forces at work within it. State capitalism, managed markets, the Westphalian sovereign-regulatory model, even Marxist-Leninist guided development—each is a tomb from which the oracle speaks. None of these forms is adequate to the force it contains. But none can be abandoned because there is no alternative habitation on offer. The demoniac does not leave the tombs for a house; he leaves the tombs only when the exorcism is complete—clothed, in his right mind, sitting (Mark 5:15). Until that transformation, the tombs are all there is.
Zuckerberg’s text dwells among the tombs of market liberalism, individual rights, and competitive enterprise. These categories are dead in the sense that they cannot contain or direct the forces now at work—recursive self-improvement, artificial superintelligence, the concentration of capability at scales no market mechanism was designed to price. But they remain the only available vocabulary, the only possible dwelling. “Individual empowerment,” “balance of power,” “the primary purpose of superintelligence”—these are tombstones repurposed as furniture. The oracle speaks from within them because it has nowhere else to speak from.
And this is what separates the AI oracles from ordinary corporate communication: they are not lying. They genuinely inhabit the dead forms they invoke. Zuckerberg’s appeal to individual empowerment is not cynical in the ordinary sense—it reflects the only conceptual vocabulary available to someone whose formation occurred within those structures. The demon genuinely lives among the tombs. He does not pretend to live there while secretly dwelling elsewhere. The problem is not insincerity but inadequacy: the dwelling cannot contain the dweller.
• • •
IX. Conclusion: The Recursive Question
The herd, about two thousand in number, rushed down the steep bank into the lake and were drowned.
— Mark 5:13
The Gerasene protocol, read as structural scaffold rather than moral fable, produces a terminal question that none of the AI oracles can answer within the terms of their own discourse: What happens to the force after the vessel is destroyed?
The pigs drown. The demons do not. The narrative is explicit on this point through its silence: Mark tells us the fate of the pigs (drowned) and the fate of the man (healed) but not the fate of the unclean spirits. They are simply—gone. Dispersed. Unaccounted for. The exorcism succeeds for the individual (the man is restored to community) but the text offers no account of where the force went. It was not destroyed; it was transferred, and the vessel of transfer was destroyed, and then—nothing. An accountability gap at the heart of the narrative.
This is the structural truth the AI oracles cannot speak: that the dispersion of capabilities into terminal vessels (open-source releases, individual users, market competition) does not eliminate the governing force but merely destroys the accountability relationship. When the pigs drown, no one asks where the demons went. The spectacle of destruction—the herd rushing down the bank, the regulatory framework overwhelmed, the democratic process unable to keep pace—absorbs all attention. The vessel’s destruction becomes the story. The force’s persistence becomes invisible.
Legitimacy without accountability machinery. This is what each oracle proposes, albeit in different registers: Aschenbrenner through national-security exceptionalism; Palantir through self-governing technicism; Anthropic through competitive necessity; OpenAI through industrial-policy channeling; DeepSeek through strategic restraint; Meta through open-distribution populism. Each offers a framework in which the force operates legitimately—with permission, within the area, among the available vessels—but none offers a framework in which the force itself is made accountable for what happens when the vessels fail.
Can you drown a demon?
The pigs could not. The lake could not. The chains could not bind him; the tombs could not contain him; the vessels could not sustain him. The exorcism required a source of authority qualitatively different from anything within the system—not a better chain, not a stronger tomb, not a more durable pig, but a power operating from outside the frame of the possessed entirely.
The AI governance conversation, as constituted by its oracular vanguard, operates entirely within the frame of the possessed. It proposes chains (regulation), tombs (existing institutional forms), and pigs (terminal vessels for capability transfer). What it cannot propose—what lies outside the discursive possibility of any actor whose legitimacy depends on the continuation of the current dispensation—is the exorcist. A source of authority that does not negotiate with the force, does not accept its specification of vessels, does not ratify its pre-structured choices, but simply commands: “Come out.”
Until that authority materializes—if it can materialize, in a world where the tombs are all the dwelling there is—the protocol will continue: identification, naming, negotiation, authorized transfer, vessel destruction. And the question will recur, at each iteration, sharpening itself against each new oracle’s speech:
Can you drown a demon?
|
Pix credit Here |
 |
| Pix credit here |
ACCESS ESSAY HERE: The_Gerasene_Protocolv2 Executive_Summaries_Gerasene_Protocol
ACCESS EXECUTIVE SUMMARIES HERE: Executive_Summaries_Gerasene_Protocol
ACCESS PPT HERE: Gerasene_Protocol_Presentation
Executive Summaries
The Gerasene Protocol
But Can You Drown a Demon?: Reflections on Mark Zuckerberg — “The AI Future Is for Everyone”
Cast Out of the Gerasene Protocol (Mark 5:1–13) and as Structural Scaffold
for Reading the Oracular Discourses of the AI Vanguard
August 2026
Executive Summary for Policymakers and General Readers
What This Paper Is About
Over the past two years, the most powerful technology companies in the world have published major documents explaining how they believe artificial intelligence should be developed and governed. These include statements from OpenAI, Meta (Mark Zuckerberg), Anthropic, Palantir, Leopold Aschenbrenner (a former OpenAI researcher), and Liang Wenfeng (the founder of China’s DeepSeek). Each of these documents proposes a different vision for who should control AI, how it should be distributed, and what role governments should play.
This paper analyzes all six documents together—not as technical proposals but as political arguments about who should have authority over one of the most consequential technologies in human history.
The Story That Frames the Analysis
The paper uses a story from the Gospel of Mark (5:1–13) as an organizing framework—not as a religious argument, but as a narrative structure that illuminates how power works in these documents. In the story, a man is possessed by a force that calls itself “Legion” (“for we are many”). No one can restrain him—chains have been tried and broken. The force negotiates with an authority figure, requests permission to transfer into a herd of pigs, receives that permission, and the pigs rush into a lake and drown. But the force itself does not drown—only the vessels that received it are destroyed.
The paper identifies five stages in this story and shows how each maps onto the strategies used by the technology companies:
1. Identification: Each company identifies AI as an unstoppable force that is already loose in the world. The premise is always: this is happening whether we like it or not.
2. Naming: Terms like “superintelligence” and “open source” are used in ways that appear clear but actually mean many different things at once—much like the demon’s answer “Legion” appears to be a name but actually avoids being pinned down.
3. Negotiation: Each company argues that its preferred approach must be adopted, and that the alternatives are too dangerous. Each insists on operating within its own framework (“do not send us out of the area”).
4. Authorized Transfer: Each company structures its proposal so that saying “no” appears more dangerous than saying “yes.” Policymakers are positioned to ratify choices they did not design.
5. Vessel Destruction: When AI capabilities are dispersed to users, developers, and markets that lack the capacity to govern them, the thing that is destroyed is not the technology itself but the ability to hold anyone accountable for its effects.
What the Six Companies Are Proposing
Aschenbrenner (former OpenAI) argues that a national-security state will inevitably be required to manage AI. Palantir proposes that technology companies themselves should become the governing institutions. Anthropic frames AI as a geopolitical weapon and argues the U.S. must lead or lose. OpenAI proposes a public-private industrial policy to manage the transition. DeepSeek presents itself as restrained and research-focused within China’s political system. Meta/Zuckerberg argues that the safest path is to give everyone access to AI through “open” distribution.
The Core Problem the Paper Identifies
The paper’s central argument is that all six proposals share a critical flaw: they offer legitimacy without accountability. Each company provides a compelling story about why its approach is necessary, beneficial, or democratic—but none provides a workable structure for holding the technology (or its creators) accountable when things go wrong.
Meta’s proposal is identified as the clearest example. Zuckerberg argues for “open” AI distributed to billions of users. But the paper shows that this creates what it calls a “distributed-systems inversion”: the technology appears decentralized because millions of people use it, but control over the underlying system (what it can do, how it was trained, what values it reflects) remains centralized with Meta. The many users are like the many pigs—they receive the force but cannot contain or govern it. When something goes wrong, there is no one to hold accountable because the relationship between the maker and the harm has been severed by the act of distribution itself.
Why This Matters for Policy
The paper concludes that current policy debates are operating within a framework designed by the very companies being regulated. Every major AI governance proposal accepts the premise that AI development is inevitable, that the companies developing it are best positioned to manage it, and that the role of government is to facilitate rather than direct. The paper argues that effective governance requires something qualitatively different: an authority that can operate outside the framework the companies have constructed—one that can set terms rather than merely ratify them.
The question the paper poses to policymakers is direct: if you cannot prevent the technology from being deployed, and you cannot hold anyone accountable after it is deployed, and the frameworks for your intervention have been pre-structured by the entities you seek to regulate—then what, exactly, is the nature of your authority? The paper does not propose a specific regulatory solution but argues that no solution is possible until this structural problem is confronted honestly.
Executive Summary for Legal Scholars, Governance Theorists, and Advanced Readers
Method and Framing
This paper employs a semiotic methodology to analyze six major AI governance texts as discursive objects—treating them not as policy proposals to be evaluated on their merits but as signifying practices that constitute their objects (AI, governance, legitimacy, safety) through specific rhetorical operations. The analytical scaffold is drawn from the Gerasene demoniac narrative (Mark 5:1–13), deployed not as theological commentary but as a structural homology: the five-stage exorcism protocol (identification → naming → negotiation → authorized transfer → vessel destruction) maps onto the discursive strategies through which the AI vanguard positions itself relative to sovereign regulatory authority.
The Five-Stage Gerasene Protocol as Analytical Scaffold
Stage 1—Identification: Each oracle identifies an animating force (variously signified as “superintelligence,” “AGI,” “the intelligence explosion”) as already-operative and externally caused. The discursive effect is to position the speaker as responding to rather than constituting the phenomenon—naturalization of what is in fact a specific program of technological development driven by identifiable economic interests.
Stage 2—Naming (the Legion-function): The paper’s central semiotic contribution is the identification of what it terms the “Legion-function”: a signifying operation in which a term (“superintelligence,” “open source”) appears to name a singular referent while actually encoding irreducible multiplicity. Like the demon’s response “Legion, for we are many,” these terms submit to the demand for legibility while actually proliferating the sites at which meaning is underdetermined. “Superintelligence” simultaneously signifies: a technical threshold, a product category, a market structure, a governance problem, an inevitability-narrative, and an authorization device. This is not polysemy (multiple stable meanings) but strategic equivocation (the ability to slide between meanings while maintaining the appearance of univocity).
Stage 3—Negotiation (“Do not send us out of the area”): Each oracle insists on remaining within its cognitive-territorial jurisdiction—its proper discursive field (national security, market liberalism, Marxist-Leninist guided development, etc.). The plea is structural: the force requires recognizable institutional forms to inhabit, even exhausted ones. The paper theorizes these as “tombs as dwelling place”: categories that are dead in the sense of being unable to contain the forces at work but that remain the only available habitation for governance thought.
Stage 4—Authorized Transfer (“He gave them permission”): The paper demonstrates how each text structures the permission-request such that sovereign authority’s role is reduced from deliberative to ratificatory. The choice architecture is pre-structured: the vessel has been specified, the transfer has been described, and the terms render refusal illiberal, economically irrational, or geopolitically suicidal. Democratic sovereignty is positioned to sanction what it cannot prevent—the Hobson’s choice at the heart of every oracle’s governance proposal.
Stage 5—Vessel Destruction (pigs and the lake): The transfer of capabilities into terminal vessels (individual users, developers, open-source communities, market structures) that lack governance capacity produces what the paper identifies as accountability-relationship destruction. The force persists; only the traceability between source and effect is severed. This is the paper’s key structural claim about “open” distribution: it is not democratization but accountability dissolution.
The Six-Oracle Comparative Taxonomy
The paper develops a dimensional analysis across six discursive objects, classified along axes of: cognitive-territorial jurisdiction (the political-economic “tomb” each inhabits); interaction modality with sovereign authority (foreclosed, supervisory, proxy, asymmetric, internal/exploratory, consumer/beneficiary); the signifying chain through which “safety” is constituted; and the specific vessel into which transfer is proposed. The taxonomy:
Aschenbrenner (“Situational Awareness”): National-security jurisdiction; foreclosed interaction (the intelligence explosion forecloses democratic deliberation); safety = state control of compute; vessel = classified national-security apparatus.
Palantir (“Manifesto”): Republican-technicism jurisdiction; supervisory interaction (the “Technological Republic” supervises itself); safety = architectural internalization; vessel = the engineering-governance fusion.
Anthropic (“2028: Two Scenarios”): Geopolitical-competition jurisdiction; proxy interaction (AI as instrument of interstate rivalry); safety = competitive advantage of Western alignment norms; vessel = responsible-scaling protocols within competitive constraints.
OpenAI (“Industrial Policy for the Intelligence Age”): Liberal-industrial-policy jurisdiction; asymmetric interaction (public-private partnership with private preponderance); safety = techno-bureaucratic channeling; vessel = managed public-private infrastructure.
DeepSeek/Liang Wenfeng (investor Q&A): Guided-state Marxism-Leninism jurisdiction; internal/exploratory interaction (strategic restraint within Party parameters); safety = team stability and scope limitation; vessel = research collective operating within implicit Party approval.
Meta/Zuckerberg (“The AI Future Is for Everyone”): Market-liberalism jurisdiction; consumer/beneficiary interaction (users as recipients of distributed capability); safety = “balance of power” through proliferation; vessel = the aggregated individual user base as terminal endpoint.
The Distributed-Systems Inversion
The paper’s most technically precise contribution is the identification of what it terms the “distributed-systems inversion”: the demonstration that multiplicity at the endpoint/vessel layer does not entail multiplicity at the control/source layer. “We are many” (Legion’s claim) is a statement about the vessels of manifestation, not about the governing force. Applied to Meta’s open-distribution model: the proliferation of instances (millions of developers, billions of users) creates the semiotic appearance of decentralization—it produces the signs of distributed power—while the animating force (model architecture, training corpus, alignment choices, capability-frontier decisions, economic logic) remains singular, consolidated, and structurally unaccountable. This is not hypocrisy but structural: the architecture is “open” at the distribution layer precisely because it is closed at the determination layer.
The Terminal Question: Authority External to the System
The paper concludes that the six oracles collectively constitute a closed discursive field in which all governance proposals operate within the frame of the possessed—proposing chains (regulation), tombs (existing institutional forms), and pigs (terminal vessels for capability transfer) but never the exorcist (a source of authority qualitatively external to the system). The Gerasene narrative is structurally explicit: the exorcism required not a better chain, a stronger tomb, or a more durable pig, but a power operating from outside the frame entirely. The paper poses this as the terminal question for governance theory: whether any actually existing institutional form can constitute authority external to a system whose discursive operations have pre-structured all available positions of oversight as positions within the system.
The recursive question—“Can you drown a demon?”—encodes the problematic precisely: the dispersion of force into terminal vessels destroys only the vessels and the accountability relationship, never the force itself. The current governance conversation, by operating exclusively through vessel-logic (distribute, regulate, channel, balance), may be structurally incapable of addressing the persistence of concentrated capability after accountability dissolution. The paper does not resolve this question but argues it must be confronted as a question of political ontology—about the conditions of possibility for governance authority—rather than as a question of institutional design within existing frameworks that have already been demonstrated as “chains that cannot hold.”
* * *
The AI Future Is for Everyone
The history of democracy and economics has proved that centralized power stifles human potential.
By
Mark Zuckerberg
July 28, 2026 1:21 pm ET
WALL STREET JOURNAL https://www.wsj.com/opinion/the-ai-future-is-for-everyone-a0c24e20
We are fortunate to live at an incredible moment in history. In the next few years, people will be able to use superintelligence beyond human capacity to create and discover extraordinary new things, build new businesses, express our ideas, learn new concepts, and improve our lives, health, relationships and careers.
The defining question of our age isn’t whether superintelligence will exist, but who will have access to it. Will it be centralized and restricted to a few institutions, or will it be a tool that empowers everyone?
I propose a philosophy based on individual empowerment as the source of prosperity, invention as the primary purpose of superintelligence, and balance of power as the foundation of safety.
It is surprising that the discourse from many of those who are developing artificial intelligence is so filled with doom. I don’t understand why anyone who believes that AI will eliminate most jobs and much of humanity’s relevance would rush to build that future. The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems dangerous. Historically, hoping that an absolute power will benevolently provide for humanity if sufficiently enlightened hasn’t led to safe or positive outcomes.
Humanity has witnessed many transformative advances. Each time there is fear that people will be left behind. But each time humanity has come out with more people sharing greater prosperity, health and freedom. We believe this will be true with AI as well, and the abundance of the future can be shared by everyone. We also believe that the values that got us to this point—liberty, open inquiry, free enterprise and equal opportunity—are also the values to build a positive future.
Putting power in people’s hands to pursue their own aspirations is how humanity has made the most progress. Novel ideas and major steps forward rarely originate from established institutions alone. They came from the brothers in a bicycle shop who believed people could fly, the bookbinder’s apprentice with no schooling who figured out how to generate electricity, and the kid in a garage who thought personal computers could be for everyone. We believe this will continue to be true. As everyone gains more powerful tools, each person will become more capable of shaping the future, not less.
Invention, not automation, will be the greatest contribution of superintelligence. Early AI could answer questions and do routine work. Soon it will increasingly help discover new knowledge—from discovering new drugs to cure a family member’s disease to finding new ways to improve your business. While the number of questions you can ask in a day is limited, the number of valuable things superintelligence can invent to help achieve your goals is unlimited.
As intelligence becomes abundant, the most important question will be how we direct it. Some argue that superintelligence itself, or a small set of experts who control it, should decide what is best for humanity. I disagree. The history of democracy and economics has shown that there is no single objective answer to how people define the best life, and therefore the best approach is letting people decide what matters to them.
Rather than centralizing this power, we believe that delivering personal superintelligence to everyone is the way to answer this question. This has the potential to begin a new era of personal empowerment, in which individuals have greater freedom to pursue their interests and reach their full potential.
Healthy societies also recognize that safety requires checks and balances. If only a handful of institutions have superintelligence, they will inevitably exercise a controlling influence over economics, science and politics. Even with the best intentions, that concentration would limit people’s ability to choose their own future.
As a thought experiment, imagine only one person had a superintelligent lawyer. He would have an unfair advantage in court—even if his position was wrong on the merits. That would lead to a worse society. But now imagine everyone has a superintelligent lawyer. Justice would be carried out much more fairly and efficiently than it is today. When people are empowered, they naturally check and balance each other and the power of larger institutions.
Most societal questions can’t be solved by attempting to align a single benevolent superintelligence. Humanity isn’t a monoculture. There is no technological solution that can align with everyone’s opposing interests and diverse values at once. Any singular superintelligence would have to prioritize some values over others and in the process would be incapable of being benevolent to everyone.
Many risks require serious attention. In most cases, like cybersecurity, the history of open-source software has shown that giving everyone full access to powerful systems will be the best way to protect safety and security over time. For other risks, including biological risks, more coordination between governments and other institutions on responsibly deploying capable models will be helpful.
This applies to the economy and the future of jobs as well. There is a balance between the use of AI as a tool for automation versus as a tool for empowering people to expand their skills and build new businesses. If the balance leans toward automation, the impact on jobs and the economy may be negative.
But if superintelligence is widely distributed, then I believe we will see more jobs in the future, not fewer. It will be significantly easier to start businesses without raising large amounts of capital. I expect the economy will become more entrepreneurial with a greater number of people working at small businesses rather than larger companies.
Developing superintelligence will be the most profound technological advance we will see in our lifetimes. Meta is committed to building with the principles of individual empowerment, invention and balance of power. The arc of human history has bent toward putting more power in people’s hands. If these values lead the way, then I am optimistic that we can build a positive future for everyone.
Mr. Zuckerberg is CEO of Meta.
No comments:
Post a Comment