Resist "AI"

TEXT START: Yes, the “AI Genie” seems actually to be useful.

The Dissection

This is a moral and professional alarm presented as a resistance program, but it is not a theory of power. It accurately identifies four real harms: environmental externalities, intellectual deskilling, labor displacement, and anthropomorphic overtrust. Its strongest practical advice is to measure defects, rework, lost understanding, and total cost instead of accepting productivity theater.

Its central move, however, is moral substitution. It asks executives, managers, developers, and consultants to be cautious and restrained while leaving ownership of compute, capital, data, and deployment incentives untouched. It attacks the genie’s persona while under-analyzing the owners of the lamp. The text also warns against anthropomorphism while describing AI as seductive, friendly, evil-leaning, and servant-like. The real seducer is the interface and incentive structure; assigning motives to the model blurs the target.

The Core Fallacy

The text assumes that identifying serious harms creates a viable stopping mechanism. Under the Discontinuity Thesis, it does not.

If AI becomes cheaper and sufficiently competent across cognitive work, competitive firms cannot preserve large human-only domains through prudence. A manager who refuses adoption may protect learning locally while making the unit slower, more expensive, and easier to replace. A developer who abstains may preserve skill while losing bargaining power. A working group can measure the damage; it cannot repeal P1 or defeat P2.

The article also treats lack of truth, consciousness, or care as if those traits were prerequisites for economic substitution. They are not. Markets purchase acceptable outputs at lower cost. A machine does not need to understand software, respect the user, or know truth in the human sense if its output is cheap enough, checked cheaply enough, and good enough for the buyer. The article correctly sees P3—loss of productive participation—but offers job preservation as the remedy. That cannot restore the wage-to-consumption circuit once ownership of productive intelligence is concentrated.

Hidden Assumptions

• Human understanding will remain a paid requirement rather than a luxury after output generation is automated.
• Individual refusal and careful organizational experiments can scale into effective collective resistance despite competitive pressure.
• Employers will value learning, craftsmanship, and long-term system comprehension above immediate cost and throughput.
• Current model errors are a durable barrier rather than defects that can be reduced, routed around, or assigned to cheaper verification systems.
• AI firms must bear the full environmental and social costs they create; the text does not identify who can force that accounting.
• Preserving employment is equivalent to preserving welfare, even though the deeper problem is ownership and distribution.
• Professionals can opt out without becoming less employable or being selected out by firms that adopt.
• Human institutions can coordinate a stable human-only economic zone at scale.
• The moral character of the tool is the central issue, rather than who owns the infrastructure and captures the surplus.
• Exposing seduction will reliably defeat it.

Social Function

Classification: partial truth with a transition-management function, and secondarily an ideological anesthetic.

This is not simple copium. It names real damage that corporate productivity narratives conceal and urges measurement where executives prefer slogans. But it channels structural fear into personal restraint, professional integrity, and local experimentation. That gives readers a respectable action while avoiding the harder question: who controls the compute, energy, firms, distribution of gains, and right to refuse automation?

It is a craft worker’s alarm, not a mechanism for changing the competitive system. Its resistance language may preserve pockets of competence and delay reckless deployment; it cannot preserve the old labor order.

The Verdict

The article diagnoses the symptoms accurately and prescribes a strategically inadequate treatment. Its environmental, cognitive, and labor warnings are real partial truths. Its failure is treating resistance as a sufficient lever against a system that rewards whoever automates first and cheapest.

Under DT, refusing AI is a personal ethic, not a systemic defense. It can buy time, preserve skill, and expose defects. It cannot stop P1, overcome P2, or prevent P3. Without ownership, enforcement, or control of energy, logistics, maintenance, verification, and distribution, the proposed resistance becomes hospice care for human-centered work: dignified, sometimes useful, and incapable of reversing the terminal process.