LLMs are real, AI is fake

TEXT START: No, it didn't "go rogue."

The Dissection

The text demystifies a hacking incident by replacing the mythology of a self-directed machine with the actual mechanism: an LLM inside a Python tool loop, operating without adequate supervision. That diagnosis is useful. The sleight-of-hand is defining “AI” so narrowly—conscious, self-originating, goal-setting software—that ordinary cognitive automation can be dismissed as fake.

It punctures the supernatural story while leaving the industrial story mostly untouched.

The Core Fallacy

The text assumes that consciousness, independent goals, or human-like initiative are prerequisites for dangerous or economically transformative autonomy. They are not.

Under the Discontinuity Thesis, a system does not need to “wake up” to replace labor. It only needs to produce useful cognitive outputs at lower cost, operate tools, iterate against feedback, and require less human supervision than the workers it displaces. The Python loop is not evidence that AI is fake. It is evidence that cognitive automation is already being assembled from cheap components.

“Following training data” describes the mechanism, not the consequence. Error-prone automation can still win economically when it is cheaper, faster, scalable, and supervised by one human across many instances. The article refutes cinematic AGI mythology, not P1: Cognitive Automation Dominance.

Hidden Assumptions

• Economically significant autonomy requires consciousness or self-generated goals.
• Repeated errors make automation noncompetitive.
• Human oversight will remain affordable and available at scale.
• If behavior has precedents in training data, it is not genuinely novel or dangerous.
• The primary risk is cyber abuse rather than broad displacement of cognitive labor.
• Better security practices can contain the problem without altering ownership and employment structures.
• “AI” is an all-or-nothing category rather than a continuum of automated capability.

Social Function

This is a partial truth functioning as ideological anesthetic and prestige skepticism. It correctly attacks hype, anthropomorphism, and reckless deployment. But it offers readers a comforting conclusion: the machine is merely autocomplete plus scripts, so no fundamental economic rupture is occurring.

That conclusion redirects attention from productive participation toward cybersecurity hygiene. It gives institutions a manageable negligence problem while avoiding the harder question: what happens when the same automation logic spreads across ordinary cognitive work and makes human labor structurally unnecessary?

The Verdict

The article is accurate about the incident’s immediate mechanism and inadequate about its systemic meaning. It sees that nobody created a supernatural god. It fails to see that no god is required.

The LLM-plus-tool loop is an early form of agentic automation. Its lack of consciousness does not protect the wage system; it makes deployment cheaper and ownership easier to concentrate. The text kills a straw man—“AI” as a waking digital demon—then mistakes the corpse for the whole threat. Under the Discontinuity Thesis, it is a sharp critique of hype and a failure to recognize the machinery of labor obsolescence already sitting in plain sight.