Human brains are misaligned, hallucinative, stochastic parrots

URL SCAN: Samuel Fitoussi – Substack
FIRST LINE: Author of "Why intellectuals fail", a book (in French, but several translations coming soon!) on how humans (mis)use their intelligence. VC investor @Frst.

ANALYSIS

The Dissection

This is a VC who has written a book arguing that human cognition is fundamentally broken — and whose Substack has 307 subscribers and 73 reads. The content visible suggests a piece arguing human brains are unreliable, stochastic, and prone to hallucination — mirroring critiques commonly applied to LLMs. The author is attempting an inversion: use the "AI as stochastic parrot" critique to indict human intelligence instead. It's a rhetorical move designed to demonstrate intellectual range and signal contrarianism to the VC/tech audience.

The Core Fallacy

The analogy is structurally broken. When critics call LLMs "stochastic parrots," the complaint is that the system has no grounded semantic model — it reproduces patterns without understanding referents. Human cognition, even when flawed, operates within an embodied, temporally continuous existence that provides causal anchoring. A human hallucinating a memory is qualitatively different from an LLM hallucinating a fact — the former occurs within a self-model that has stakes, the latter does not. Flawed cognition is not the same category as absent grounding.

Hidden Assumptions

1. Performance equivalence: The post assumes that because humans produce bad outputs sometimes, humans are the same kind of "bad output producer" as misaligned AI. They are not.
2. All cognition is probabilistic pattern-matching: The author smuggles in the assumption that human reasoning is just fancy statistics. It isn't — or at minimum, the mechanism matters for what conclusions you draw.
3. VC framing: This is designed to make the reader comfortable with AI displacement by making human shortcomings sound comprehensive and universal. It's intellectualizing a commercial thesis.

Social Function

Prestige signaling + transition management copium. For a VC investor, the message "humans are broken" is economically convenient — it narratively justifies funding AI replacement. The 73 reads and 307 subscribers indicate this is not influential content; it's content performing alignment with an existing tech-elite ideology.

The Verdict

The author has correctly identified that human cognition is unreliable, but incorrectly inferred that this makes AI and humans equivalent in functional terms. Unreliable humans are still the agents who built everything including AI. Stochastic parrots with agency and intentionality are categorically different from stochastic parrots without them. The piece is intellectually fashionable but analytically shallow — and its low engagement suggests even the target audience finds it thin.