Superintelligence is a dragon - Platformer

TEXT START: This is a column about AI.

The Dissection

The text is an attack on Zuckerberg’s manifesto disguised as a dragon parable. Its real targets are threefold: Meta’s commercial hypocrisy, the fantasy that mass distribution automatically produces safety, and the industry’s refusal to confront loss of control.

Its strongest move is separating access from safety. Giving everyone powerful systems does not create checks and balances if the underlying compute, capital, energy, infrastructure, and deployment channels remain concentrated. “Democratization” can simply mean wider access to tools still owned by a small number of Sovereigns.

But the article spends its analytical ammunition on catastrophic control failures and regulatory caution while treating employment destruction as an unresolved side effect. That is the central omission.

The Core Fallacy

The text mistakes the dragon’s danger for the economy’s cause of death.

Under the Discontinuity Thesis, superintelligence does not need to escape, launch a cyberattack, or engineer a pandemic to terminate the post-WWII order. It only needs to outperform human cognitive labor at durable scale. That severs the mass employment → wage → consumption circuit:

• P1: cognitive work becomes automatable.
• P2: institutions cannot preserve stable human-only economic domains at scale.
• P3: most people lose access to economically necessary labor.

A safely controlled superintelligence can therefore still destroy mass labor’s economic role. Catastrophic misalignment is an additional extinction pathway, not the necessary mechanism of systemic obsolescence.

The article also treats “net effect on jobs” as the relevant measure. That is an accounting trick. Aggregate invention and output can rise while the majority lose wages, bargaining power, and productive necessity. New tools do not imply a new mass employment circuit.

Hidden Assumptions

• Technical control can be achieved before competitive deployment pressure overwhelms it.
• Regulatory institutions can impose durable restraint despite national rivalry, corporate lobbying, and capture.
• Slowing development buys more than time; it can reverse the underlying automation dynamic.
• Wider model access produces countervailing power, even though access is not ownership or control of AI capital.
• New invention will generate enough economically necessary human roles to offset automation.
• Catastrophic risks can be detected, contained, and iterated away before an irreversible event.

These assumptions collide with the thesis’s coordination-impossibility constraint. Legal and institutional friction can delay the break. It cannot restore the old labor equilibrium once superior machine cognition becomes broadly deployable.

Social Function

Primary classification: partial truth and transition management.

The article punctures boosterist propaganda, exposes Meta’s self-serving policy agenda, and correctly argues that “give everyone the dragon” is not a safety strategy. But by centering spectacular control disasters, it makes the quieter and more certain dispossession of labor look secondary. It is a warning label attached to an industrial demolition project, not a complete account of who loses economic sovereignty when the machines work.

It also performs elite accountability without offering a structural answer to ownership. Slowdown, oversight, and cautious deployment may reduce immediate hazard and delay concentration. They do not establish Sovereign status for ordinary people, nor do they preserve mass productive participation.

The Verdict

This is a sharp takedown of Zuckerberg’s manifesto and a credible warning about control risk. It is not a complete autopsy of the economic order.

Zuckerberg is wrong that universal distribution is inherently safe. The article is wrong, or at least radically incomplete, if it implies that control is the decisive question. The decisive question is whether humans remain economically necessary. If the answer is no, post-WWII capitalism is already structurally dead—even if the dragon remains perfectly leashed.